gre-ilya/libuv
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
master
libuv/test/test-metrics.c

394 lines
11 KiB
C
Raw Permalink Normal View History

nfci: address some style nits PR-URL: https://github.com/libuv/libuv/pull/2954 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
2020-08-21 21:25:42 +00:00
/* Copyright libuv project contributors. All rights reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal in the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
* IN THE SOFTWARE.
*/
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
#include "uv.h"
#include "task.h"
#include <string.h> /* memset */
nfci: address some style nits PR-URL: https://github.com/libuv/libuv/pull/2954 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
2020-08-21 21:25:42 +00:00
#define UV_NS_TO_MS 1000000
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
typedef struct {
uv_fs_t open_req;
uv_fs_t write_req;
uv_fs_t close_req;
} fs_reqs_t;
static uint64_t last_events_count;
static char test_buf[] = "test-buffer\n";
static fs_reqs_t fs_reqs;
src: fix events/events_waiting metrics counter (#3957) The worker pool calls all callbacks locally within the queue. So the value of nevents doesn't properly reflect that case. Increase the number of events directly from the worker pool's callback to correct this. In order to properly determine if the events_waiting counter needs to be incremented, store the timeout value at the time the event provider was called.
2023-04-17 18:48:39 +00:00
static int pool_events_counter;
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
static void timer_spin_cb(uv_timer_t* handle) {
uint64_t t;
(*(int*) handle->data)++;
t = uv_hrtime();
/* Spin for 500 ms to spin loop time out of the delta check. */
nfci: address some style nits PR-URL: https://github.com/libuv/libuv/pull/2954 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
2020-08-21 21:25:42 +00:00
while (uv_hrtime() - t < 600 * UV_NS_TO_MS) { }
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
}
TEST_IMPL(metrics_idle_time) {
build: add CI for OpenBSD and FreeBSD (#3548) Fixes: https://github.com/libuv/libuv/issues/3510
2023-01-21 12:52:36 +00:00
#if defined(__OpenBSD__)
RETURN_SKIP("Test does not currently work in OpenBSD");
#endif
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
const uint64_t timeout = 1000;
uv_timer_t timer;
uint64_t idle_time;
int cntr;
cntr = 0;
timer.data = &cntr;
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_loop_configure(uv_default_loop(), UV_METRICS_IDLE_TIME));
ASSERT_OK(uv_timer_init(uv_default_loop(), &timer));
ASSERT_OK(uv_timer_start(&timer, timer_spin_cb, timeout, 0));
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_run(uv_default_loop(), UV_RUN_DEFAULT));
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
ASSERT_GT(cntr, 0);
idle_time = uv_metrics_idle_time(uv_default_loop());
/* Permissive check that the idle time matches within the timeout ±500 ms. */
build: add CI for OpenBSD and FreeBSD (#3548) Fixes: https://github.com/libuv/libuv/issues/3510
2023-01-21 12:52:36 +00:00
ASSERT_LE(idle_time, (timeout + 500) * UV_NS_TO_MS);
ASSERT_GE(idle_time, (timeout - 500) * UV_NS_TO_MS);
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
test: silence more valgrind warnings (#3917) Pass the loop to MAKE_VALGRIND_HAPPY() so it's explicit on which loop needs to be cleaned up. Since it asserts on uv_loop_close(), need to remove a couple of those that were being done before the call. Cleanup where loop was assigned, so the entire test either uses loop or uv_default_loop(). Not both. Also take care of any reqs that may have been left uncleaned.
2023-03-12 13:59:00 +00:00
MAKE_VALGRIND_HAPPY(uv_default_loop());
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
return 0;
}
static void metrics_routine_cb(void* arg) {
const uint64_t timeout = 1000;
uv_loop_t loop;
uv_timer_t timer;
uint64_t idle_time;
int cntr;
cntr = 0;
timer.data = &cntr;
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_loop_init(&loop));
ASSERT_OK(uv_loop_configure(&loop, UV_METRICS_IDLE_TIME));
ASSERT_OK(uv_timer_init(&loop, &timer));
ASSERT_OK(uv_timer_start(&timer, timer_spin_cb, timeout, 0));
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_run(&loop, UV_RUN_DEFAULT));
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
ASSERT_GT(cntr, 0);
idle_time = uv_metrics_idle_time(&loop);
/* Only checking that idle time is greater than the lower bound since there
* may have been thread contention, causing the event loop to be delayed in
* the idle phase longer than expected.
*/
nfci: address some style nits PR-URL: https://github.com/libuv/libuv/pull/2954 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
2020-08-21 21:25:42 +00:00
ASSERT_GE(idle_time, (timeout - 500) * UV_NS_TO_MS);
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
close_loop(&loop);
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_loop_close(&loop));
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
}
TEST_IMPL(metrics_idle_time_thread) {
uv_thread_t threads[5];
int i;
for (i = 0; i < 5; i++) {
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_thread_create(&threads[i], metrics_routine_cb, NULL));
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
}
for (i = 0; i < 5; i++) {
uv_thread_join(&threads[i]);
}
return 0;
}
static void timer_noop_cb(uv_timer_t* handle) {
(*(int*) handle->data)++;
}
TEST_IMPL(metrics_idle_time_zero) {
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
uv_metrics_t metrics;
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
uv_timer_t timer;
int cntr;
cntr = 0;
timer.data = &cntr;
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_loop_configure(uv_default_loop(), UV_METRICS_IDLE_TIME));
ASSERT_OK(uv_timer_init(uv_default_loop(), &timer));
ASSERT_OK(uv_timer_start(&timer, timer_noop_cb, 0, 0));
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_run(uv_default_loop(), UV_RUN_DEFAULT));
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
ASSERT_GT(cntr, 0);
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_metrics_idle_time(uv_default_loop()));
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_metrics_info(uv_default_loop(), &metrics));
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
ASSERT_UINT64_EQ(cntr, metrics.loop_count);
test: silence more valgrind warnings (#3917) Pass the loop to MAKE_VALGRIND_HAPPY() so it's explicit on which loop needs to be cleaned up. Since it asserts on uv_loop_close(), need to remove a couple of those that were being done before the call. Cleanup where loop was assigned, so the entire test either uses loop or uv_default_loop(). Not both. Also take care of any reqs that may have been left uncleaned.
2023-03-12 13:59:00 +00:00
MAKE_VALGRIND_HAPPY(uv_default_loop());
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
return 0;
}
static void close_cb(uv_fs_t* req) {
uv_metrics_t metrics;
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_metrics_info(uv_default_loop(), &metrics));
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
ASSERT_UINT64_EQ(3, metrics.loop_count);
ASSERT_UINT64_GT(metrics.events, last_events_count);
uv_fs_req_cleanup(req);
last_events_count = metrics.events;
}
static void write_cb(uv_fs_t* req) {
uv_metrics_t metrics;
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_metrics_info(uv_default_loop(), &metrics));
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
ASSERT_UINT64_EQ(2, metrics.loop_count);
ASSERT_UINT64_GT(metrics.events, last_events_count);
ASSERT_EQ(req->result, sizeof(test_buf));
uv_fs_req_cleanup(req);
last_events_count = metrics.events;
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_fs_close(uv_default_loop(),
&fs_reqs.close_req,
2023.11.06, Version 1.47.0 (Stable) Changes since version 1.46.0: * test: fix license blurb (Ben Noordhuis) * linux: fix harmless warn_unused_result warning (Shuduo Sang) * darwin: fix build warnings (小明) * linux: don't use io_uring on pre-5.10.186 kernels (Ben Noordhuis) * fs: fix WTF-8 decoding issue (Jameson Nash) * test: enable disabled tcp_connect6_error_fault (Ben Noordhuis) * test: enable disabled fs_link (Ben Noordhuis) * test: enable disabled spawn_same_stdout_stderr (Ben Noordhuis) * linux: handle UNAME26 personality (Ben Noordhuis) * build: move cmake_minimum_required version to 3.9 (Keith Winstein) * unix: set ipv6 scope id for link-local addresses (Ben Noordhuis) * unix: match kqueue and epoll code (Trevor Norris) * win,spawn: allow `%PATH%` to be unset (Kyle Edwards) * doc: switch to Furo, a more modern Sphinx theme (Saúl Ibarra Corretgé) * darwin: make TCP_KEEPINTVL and TCP_KEEPCNT available (小明) * win,fs: avoid winapi macro redefinition (Brad King) * linux: add missing riscv syscall numbers (michalbiesek) * doc: fix broken "Shared library" Wikipedia link (Alois Klink) * unix: get mainline kernel version in Ubuntu (Santiago Gimeno) * unix: get mainline kernel version in Debian (Ben Noordhuis) * build: fix qemu install in CI-unix workflow (Santiago Gimeno) * unix: disable io_uring close on selected kernels (Santiago Gimeno) * test: skip tests when ipv6 is not available (Santiago Gimeno) * ibmi: implement ifaddrs, getifaddrs, freeifaddrs (Abdirahim Musse) * unix: reset signal counters after fork (SmorkalovG) * win,process: avoid assert after spawning Store app (Jameson Nash) * unix: remove pread/preadv conditionals (Ben Noordhuis) * unix: remove pwrite/pwritev conditionals (Ben Noordhuis) * darwin: remove workaround for data corruption bug (Ben Noordhuis) * src: default to stream=stderr in handle printer (Ben Noordhuis) * test: switch to new-style ASSERT_EQ macros (Pleuvens) * zos: correctly get cpu model in uv_cpu_info() (jolai) * test: fix get_passwd2 on IBM i (Abdirahim Musse) * unix: don't malloc on sync uv_fs_read (Ben Noordhuis) * freebsd: get fs event path with fcntl(F_KINFO) (David Carlier) * test: switch from ASSERT_* to ASSERT_PTR_* (Pleuvens) * darwin: workaround apple pthread_cond_wait bug (Julien Roncaglia) * doc: uv_close should be called after exit callback (Pleuvens) * test: 192.0.2.0/24 is the actual -TEST-NET-1 (prubel) * unix: add back preadv/pwritev fallback (Ben Noordhuis) * unix: rename variable for consistency (Ben Noordhuis) * unix: merge read/write code into single functions (Ben Noordhuis) * doc: filename arg to uv_fs_event_cb can be NULL (Ben Noordhuis) * build,win: we need to link against shell32.lib (Per Allansson) * unix: no preadv/pwritev workaround if not needed (Jeffrey H. Johnson) * build: add CI for Windows ARM64 (build only) (Per Allansson) * linux: disable io_uring on 32 bits arm systems (Ben Noordhuis) * build: run sanitizers on macos ci (Ben Noordhuis) * misc: export WTF8 conversion utilities (Jameson Nash) * build: fix libuv.a file name for cmake (Jameson Nash) * build: add windows ubsan and clang ci (Matheus Izvekov) * win: improve accuracy of ProductName between arch (Christian Heimlich) -----BEGIN PGP SIGNATURE----- iQIzBAABCgAdFiEEz7ucqaW+r9cOKzxaeaZ8VaNnnIsFAmVJLa4ACgkQeaZ8VaNn nIvdMw//XQzU1OasKcghWZZAFmrgy8D+40Q5drSgCFryAA+WA9wXruKOeqQ3vw7+ 9Xwai97fMbxGSFhR2IwjbwD8o8DZotDTMBgSUvuyAlzdw0xB2D+eAnvxaMmYg2km R5eNb2KgmzhC7HFuPbpjWbL1mEnBU8rTSzV6yeMs2LXfLquJoopmgAjB2MXS6/Fr +WkwJx7/B/3wEhhbX5sopd6PT8UcWdbSvB/S+PU12kI0KgJeJlq0jXd5LyDaQYI6 j4XYcpgEp8B/gyiXSgso+eg1CiZQUYVMXZ5+6YA5le/s5ieJwJ2cEvo/HJa5guZ/ PBb+ulucUaupBzQFaPYwnrEzE3YHNVppBcfryxqn48Ih0Z9pNs/6EdxqggYvAyWk VA9Yi0qDFSFgD0hZSkvi/m/UEH17FLTac1DSob4EKfwAJ04spmEtiBWaL+7wQ/ca Kc819NXBaFUb+BEkN711B1CvaT6aef9wZcis42MPHjHJ19BNgb4uAa3GbmrvB2dt EodLGlbOAkKLFC29qAGeKZKhnSWfZS/HvuyqMOgONuJTj0Jm/RdwqIsvILejtj32 7m9saTkRWav4L1yvksUYT24WIk1e324gBgm9411IggdDILgbLYrwWNZb56FgMDQb cqmg0Ap6zMZ9UOf4hfqTiKZPjqqfxgSp20zCjVTy+vsCSXJnFaQ= =7yzD -----END PGP SIGNATURE----- Merge tag 'v1.47.0' into merge_1.47.0
2023-11-15 15:50:31 +00:00
(uv_os_fd_t) fs_reqs.open_req.result,
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
close_cb));
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
}
static void create_cb(uv_fs_t* req) {
uv_metrics_t metrics;
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_metrics_info(uv_default_loop(), &metrics));
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
/* Event count here is still 0 so not going to check. */
ASSERT_UINT64_EQ(1, metrics.loop_count);
ASSERT_GE(req->result, 0);
uv_fs_req_cleanup(req);
last_events_count = metrics.events;
uv_buf_t iov = uv_buf_init(test_buf, sizeof(test_buf));
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_fs_write(uv_default_loop(),
&fs_reqs.write_req,
2023.11.06, Version 1.47.0 (Stable) Changes since version 1.46.0: * test: fix license blurb (Ben Noordhuis) * linux: fix harmless warn_unused_result warning (Shuduo Sang) * darwin: fix build warnings (小明) * linux: don't use io_uring on pre-5.10.186 kernels (Ben Noordhuis) * fs: fix WTF-8 decoding issue (Jameson Nash) * test: enable disabled tcp_connect6_error_fault (Ben Noordhuis) * test: enable disabled fs_link (Ben Noordhuis) * test: enable disabled spawn_same_stdout_stderr (Ben Noordhuis) * linux: handle UNAME26 personality (Ben Noordhuis) * build: move cmake_minimum_required version to 3.9 (Keith Winstein) * unix: set ipv6 scope id for link-local addresses (Ben Noordhuis) * unix: match kqueue and epoll code (Trevor Norris) * win,spawn: allow `%PATH%` to be unset (Kyle Edwards) * doc: switch to Furo, a more modern Sphinx theme (Saúl Ibarra Corretgé) * darwin: make TCP_KEEPINTVL and TCP_KEEPCNT available (小明) * win,fs: avoid winapi macro redefinition (Brad King) * linux: add missing riscv syscall numbers (michalbiesek) * doc: fix broken "Shared library" Wikipedia link (Alois Klink) * unix: get mainline kernel version in Ubuntu (Santiago Gimeno) * unix: get mainline kernel version in Debian (Ben Noordhuis) * build: fix qemu install in CI-unix workflow (Santiago Gimeno) * unix: disable io_uring close on selected kernels (Santiago Gimeno) * test: skip tests when ipv6 is not available (Santiago Gimeno) * ibmi: implement ifaddrs, getifaddrs, freeifaddrs (Abdirahim Musse) * unix: reset signal counters after fork (SmorkalovG) * win,process: avoid assert after spawning Store app (Jameson Nash) * unix: remove pread/preadv conditionals (Ben Noordhuis) * unix: remove pwrite/pwritev conditionals (Ben Noordhuis) * darwin: remove workaround for data corruption bug (Ben Noordhuis) * src: default to stream=stderr in handle printer (Ben Noordhuis) * test: switch to new-style ASSERT_EQ macros (Pleuvens) * zos: correctly get cpu model in uv_cpu_info() (jolai) * test: fix get_passwd2 on IBM i (Abdirahim Musse) * unix: don't malloc on sync uv_fs_read (Ben Noordhuis) * freebsd: get fs event path with fcntl(F_KINFO) (David Carlier) * test: switch from ASSERT_* to ASSERT_PTR_* (Pleuvens) * darwin: workaround apple pthread_cond_wait bug (Julien Roncaglia) * doc: uv_close should be called after exit callback (Pleuvens) * test: 192.0.2.0/24 is the actual -TEST-NET-1 (prubel) * unix: add back preadv/pwritev fallback (Ben Noordhuis) * unix: rename variable for consistency (Ben Noordhuis) * unix: merge read/write code into single functions (Ben Noordhuis) * doc: filename arg to uv_fs_event_cb can be NULL (Ben Noordhuis) * build,win: we need to link against shell32.lib (Per Allansson) * unix: no preadv/pwritev workaround if not needed (Jeffrey H. Johnson) * build: add CI for Windows ARM64 (build only) (Per Allansson) * linux: disable io_uring on 32 bits arm systems (Ben Noordhuis) * build: run sanitizers on macos ci (Ben Noordhuis) * misc: export WTF8 conversion utilities (Jameson Nash) * build: fix libuv.a file name for cmake (Jameson Nash) * build: add windows ubsan and clang ci (Matheus Izvekov) * win: improve accuracy of ProductName between arch (Christian Heimlich) -----BEGIN PGP SIGNATURE----- iQIzBAABCgAdFiEEz7ucqaW+r9cOKzxaeaZ8VaNnnIsFAmVJLa4ACgkQeaZ8VaNn nIvdMw//XQzU1OasKcghWZZAFmrgy8D+40Q5drSgCFryAA+WA9wXruKOeqQ3vw7+ 9Xwai97fMbxGSFhR2IwjbwD8o8DZotDTMBgSUvuyAlzdw0xB2D+eAnvxaMmYg2km R5eNb2KgmzhC7HFuPbpjWbL1mEnBU8rTSzV6yeMs2LXfLquJoopmgAjB2MXS6/Fr +WkwJx7/B/3wEhhbX5sopd6PT8UcWdbSvB/S+PU12kI0KgJeJlq0jXd5LyDaQYI6 j4XYcpgEp8B/gyiXSgso+eg1CiZQUYVMXZ5+6YA5le/s5ieJwJ2cEvo/HJa5guZ/ PBb+ulucUaupBzQFaPYwnrEzE3YHNVppBcfryxqn48Ih0Z9pNs/6EdxqggYvAyWk VA9Yi0qDFSFgD0hZSkvi/m/UEH17FLTac1DSob4EKfwAJ04spmEtiBWaL+7wQ/ca Kc819NXBaFUb+BEkN711B1CvaT6aef9wZcis42MPHjHJ19BNgb4uAa3GbmrvB2dt EodLGlbOAkKLFC29qAGeKZKhnSWfZS/HvuyqMOgONuJTj0Jm/RdwqIsvILejtj32 7m9saTkRWav4L1yvksUYT24WIk1e324gBgm9411IggdDILgbLYrwWNZb56FgMDQb cqmg0Ap6zMZ9UOf4hfqTiKZPjqqfxgSp20zCjVTy+vsCSXJnFaQ= =7yzD -----END PGP SIGNATURE----- Merge tag 'v1.47.0' into merge_1.47.0
2023-11-15 15:50:31 +00:00
(uv_os_fd_t) req->result,
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
&iov,
1,
0,
write_cb));
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
}
static void prepare_cb(uv_prepare_t* handle) {
uv_metrics_t metrics;
uv_prepare_stop(handle);
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_metrics_info(uv_default_loop(), &metrics));
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
ASSERT_UINT64_EQ(0, metrics.loop_count);
ASSERT_UINT64_EQ(0, metrics.events);
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_fs_open(uv_default_loop(),
&fs_reqs.open_req,
"test_file",
2023.11.06, Version 1.47.0 (Stable) Changes since version 1.46.0: * test: fix license blurb (Ben Noordhuis) * linux: fix harmless warn_unused_result warning (Shuduo Sang) * darwin: fix build warnings (小明) * linux: don't use io_uring on pre-5.10.186 kernels (Ben Noordhuis) * fs: fix WTF-8 decoding issue (Jameson Nash) * test: enable disabled tcp_connect6_error_fault (Ben Noordhuis) * test: enable disabled fs_link (Ben Noordhuis) * test: enable disabled spawn_same_stdout_stderr (Ben Noordhuis) * linux: handle UNAME26 personality (Ben Noordhuis) * build: move cmake_minimum_required version to 3.9 (Keith Winstein) * unix: set ipv6 scope id for link-local addresses (Ben Noordhuis) * unix: match kqueue and epoll code (Trevor Norris) * win,spawn: allow `%PATH%` to be unset (Kyle Edwards) * doc: switch to Furo, a more modern Sphinx theme (Saúl Ibarra Corretgé) * darwin: make TCP_KEEPINTVL and TCP_KEEPCNT available (小明) * win,fs: avoid winapi macro redefinition (Brad King) * linux: add missing riscv syscall numbers (michalbiesek) * doc: fix broken "Shared library" Wikipedia link (Alois Klink) * unix: get mainline kernel version in Ubuntu (Santiago Gimeno) * unix: get mainline kernel version in Debian (Ben Noordhuis) * build: fix qemu install in CI-unix workflow (Santiago Gimeno) * unix: disable io_uring close on selected kernels (Santiago Gimeno) * test: skip tests when ipv6 is not available (Santiago Gimeno) * ibmi: implement ifaddrs, getifaddrs, freeifaddrs (Abdirahim Musse) * unix: reset signal counters after fork (SmorkalovG) * win,process: avoid assert after spawning Store app (Jameson Nash) * unix: remove pread/preadv conditionals (Ben Noordhuis) * unix: remove pwrite/pwritev conditionals (Ben Noordhuis) * darwin: remove workaround for data corruption bug (Ben Noordhuis) * src: default to stream=stderr in handle printer (Ben Noordhuis) * test: switch to new-style ASSERT_EQ macros (Pleuvens) * zos: correctly get cpu model in uv_cpu_info() (jolai) * test: fix get_passwd2 on IBM i (Abdirahim Musse) * unix: don't malloc on sync uv_fs_read (Ben Noordhuis) * freebsd: get fs event path with fcntl(F_KINFO) (David Carlier) * test: switch from ASSERT_* to ASSERT_PTR_* (Pleuvens) * darwin: workaround apple pthread_cond_wait bug (Julien Roncaglia) * doc: uv_close should be called after exit callback (Pleuvens) * test: 192.0.2.0/24 is the actual -TEST-NET-1 (prubel) * unix: add back preadv/pwritev fallback (Ben Noordhuis) * unix: rename variable for consistency (Ben Noordhuis) * unix: merge read/write code into single functions (Ben Noordhuis) * doc: filename arg to uv_fs_event_cb can be NULL (Ben Noordhuis) * build,win: we need to link against shell32.lib (Per Allansson) * unix: no preadv/pwritev workaround if not needed (Jeffrey H. Johnson) * build: add CI for Windows ARM64 (build only) (Per Allansson) * linux: disable io_uring on 32 bits arm systems (Ben Noordhuis) * build: run sanitizers on macos ci (Ben Noordhuis) * misc: export WTF8 conversion utilities (Jameson Nash) * build: fix libuv.a file name for cmake (Jameson Nash) * build: add windows ubsan and clang ci (Matheus Izvekov) * win: improve accuracy of ProductName between arch (Christian Heimlich) -----BEGIN PGP SIGNATURE----- iQIzBAABCgAdFiEEz7ucqaW+r9cOKzxaeaZ8VaNnnIsFAmVJLa4ACgkQeaZ8VaNn nIvdMw//XQzU1OasKcghWZZAFmrgy8D+40Q5drSgCFryAA+WA9wXruKOeqQ3vw7+ 9Xwai97fMbxGSFhR2IwjbwD8o8DZotDTMBgSUvuyAlzdw0xB2D+eAnvxaMmYg2km R5eNb2KgmzhC7HFuPbpjWbL1mEnBU8rTSzV6yeMs2LXfLquJoopmgAjB2MXS6/Fr +WkwJx7/B/3wEhhbX5sopd6PT8UcWdbSvB/S+PU12kI0KgJeJlq0jXd5LyDaQYI6 j4XYcpgEp8B/gyiXSgso+eg1CiZQUYVMXZ5+6YA5le/s5ieJwJ2cEvo/HJa5guZ/ PBb+ulucUaupBzQFaPYwnrEzE3YHNVppBcfryxqn48Ih0Z9pNs/6EdxqggYvAyWk VA9Yi0qDFSFgD0hZSkvi/m/UEH17FLTac1DSob4EKfwAJ04spmEtiBWaL+7wQ/ca Kc819NXBaFUb+BEkN711B1CvaT6aef9wZcis42MPHjHJ19BNgb4uAa3GbmrvB2dt EodLGlbOAkKLFC29qAGeKZKhnSWfZS/HvuyqMOgONuJTj0Jm/RdwqIsvILejtj32 7m9saTkRWav4L1yvksUYT24WIk1e324gBgm9411IggdDILgbLYrwWNZb56FgMDQb cqmg0Ap6zMZ9UOf4hfqTiKZPjqqfxgSp20zCjVTy+vsCSXJnFaQ= =7yzD -----END PGP SIGNATURE----- Merge tag 'v1.47.0' into merge_1.47.0
2023-11-15 15:50:31 +00:00
UV_FS_O_WRONLY | UV_FS_O_CREAT, S_IRUSR | S_IWUSR,
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
create_cb));
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
}
TEST_IMPL(metrics_info_check) {
uv_fs_t unlink_req;
uv_prepare_t prepare;
uv_fs_unlink(NULL, &unlink_req, "test_file", NULL);
uv_fs_req_cleanup(&unlink_req);
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_prepare_init(uv_default_loop(), &prepare));
ASSERT_OK(uv_prepare_start(&prepare, prepare_cb));
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
test: switch to new-style ASSERT_EQ macros (#4159) Switch from old-style ASSERT macro to new-style ASSERT_EQ,... macros. Using new-style macros makes it easier to debug test failures Fixes: https://github.com/libuv/libuv/issues/2974
2023-10-06 17:50:15 +00:00
ASSERT_OK(uv_run(uv_default_loop(), UV_RUN_DEFAULT));
src: add new metrics APIs (#3749) The following metrics are now always recorded and available via the new uv_metrics_info() API. * loop_count: Number of event loop iterations. * events: Total number of events processed by the event handler. * events_waiting: Total number of events waiting in the event queue when the event provider request was made. Benchmarking has shown no noticeable impact recording these metrics. PR-URL: https://github.com/libuv/libuv/pull/3749
2022-11-11 15:21:58 +00:00
uv_fs_unlink(NULL, &unlink_req, "test_file", NULL);
uv_fs_req_cleanup(&unlink_req);
test: silence more valgrind warnings (#3917) Pass the loop to MAKE_VALGRIND_HAPPY() so it's explicit on which loop needs to be cleaned up. Since it asserts on uv_loop_close(), need to remove a couple of those that were being done before the call. Cleanup where loop was assigned, so the entire test either uses loop or uv_default_loop(). Not both. Also take care of any reqs that may have been left uncleaned.
2023-03-12 13:59:00 +00:00
MAKE_VALGRIND_HAPPY(uv_default_loop());
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
return 0;
}
src: fix events/events_waiting metrics counter (#3957) The worker pool calls all callbacks locally within the queue. So the value of nevents doesn't properly reflect that case. Increase the number of events directly from the worker pool's callback to correct this. In order to properly determine if the events_waiting counter needs to be incremented, store the timeout value at the time the event provider was called.
2023-04-17 18:48:39 +00:00
static void fs_prepare_cb(uv_prepare_t* handle) {
uv_metrics_t metrics;
ASSERT_OK(uv_metrics_info(uv_default_loop(), &metrics));
if (pool_events_counter == 1)
ASSERT_EQ(metrics.events, metrics.events_waiting);
if (pool_events_counter < 7)
return;
uv_prepare_stop(handle);
pool_events_counter = -42;
}
static void fs_stat_cb(uv_fs_t* req) {
uv_fs_req_cleanup(req);
pool_events_counter++;
}
static void fs_work_cb(uv_work_t* req) {
}
static void fs_after_work_cb(uv_work_t* req, int status) {
free(req);
pool_events_counter++;
}
static void fs_write_cb(uv_fs_t* req) {
uv_work_t* work1 = malloc(sizeof(*work1));
uv_work_t* work2 = malloc(sizeof(*work2));
pool_events_counter++;
uv_fs_req_cleanup(req);
ASSERT_OK(uv_queue_work(uv_default_loop(),
work1,
fs_work_cb,
fs_after_work_cb));
ASSERT_OK(uv_queue_work(uv_default_loop(),
work2,
fs_work_cb,
fs_after_work_cb));
}
static void fs_random_cb(uv_random_t* req, int status, void* buf, size_t len) {
pool_events_counter++;
}
static void fs_addrinfo_cb(uv_getaddrinfo_t* req,
int status,
struct addrinfo* res) {
uv_freeaddrinfo(req->addrinfo);
pool_events_counter++;
}
TEST_IMPL(metrics_pool_events) {
uv_buf_t iov;
uv_fs_t open_req;
uv_fs_t stat1_req;
uv_fs_t stat2_req;
uv_fs_t unlink_req;
uv_fs_t write_req;
uv_getaddrinfo_t addrinfo_req;
uv_metrics_t metrics;
uv_prepare_t prepare;
uv_random_t random_req;
2023.05.19, Version 1.45.0 (Stable) Changes since version 1.44.2: * win: remove stdint-msvc2008.h (Ben Noordhuis) * android: remove pthread-fixes.c (Ben Noordhuis) * build: enable MSVC_RUNTIME_LIBRARY setting (自发对称破缺) * unix: switch to c11 atomics (Ben Noordhuis) * unix: don't accept() connections in a loop (Ben Noordhuis) * win: fix off-by-1 buffer overrun in uv_exepath() (Ben Noordhuis) * build: switch ci from macos-10.15 to macos-11 (Ben Noordhuis) * win: fix thread race in uv_cwd() and uv_chdir() (Ben Noordhuis) * unix,win: remove UV_HANDLE_SHUTTING flag (Santiago Gimeno) * win: support Windows 11 in uv_os_uname() (Luan Devecchi) * unix: fix uv_getrusage() ru_maxrss reporting (Ben Noordhuis) * doc: add note about offset -1 in uv_fs_read/write (Steven Schveighoffer) * test: fix musl libc.a dlerror() test expectation (Ben Noordhuis) * kqueue: DRY file descriptor deletion logic (Ben Noordhuis) * linux: teach uv_get_constrained_memory() cgroupsv2 (Ben Noordhuis) * build: upgrade qemu-user-static package (Ben Noordhuis) * linux: move epoll.c back into linux-core.c (Ben Noordhuis) * unix: remove pre-macos 10.8 compatibility hack (Ben Noordhuis) * unix,win: fix memory leak in uv_fs_scandir() (Ben Noordhuis) * build: restore qemu download logic (Ben Noordhuis) * win: fix uv__pipe_accept memory leak (number201724) * doc: update LINKS.md (Daniel) * unix: simplify atomic op in uv_tty_reset_mode() (Ben Noordhuis) * build: add LIBUV_BUILD_SHARED cmake option (Christian Clason) * linux: remove unused or obsolete syscall wrappers (Ben Noordhuis) * linux: merge files back into single file (Ben Noordhuis) * stream: process more than one write req per loop tick (ywave620) * unix,win: give thread pool threads an 8 MB stack (Ben Noordhuis) * build: add MemorySanitizer (MSAN) support (Ben Noordhuis) * doc: add uv_poll_cb status==UV_EBADF note (jensbjorgensen) * build: support AddressSanitizer on MSVC (Jameson Nash) * win,pipe: improve method of obtaining pid for ipc (number201724) * thread: add support for affinity (daomingq) * include: map ENODATA error code (Ben Noordhuis) * build: remove bashism from autogen.sh (Santiago Gimeno) * win,tcp,udp: remove "active streams" optimization (Saúl Ibarra Corretgé) * win: drop code checking for Windows XP / Server 2k3 (Saúl Ibarra Corretgé) * unix,win: fix 'sprintf' is deprecated warning (twosee) * doc: mention close_cb can be NULL (Qix) * win: optimize udp receive performance (ywave620) * win: fix an incompatible types warning (twosee) * doc: document 0 return value for free/total memory (Ben Noordhuis) * darwin: use hw.cpufrequency again for frequency info (Jameson Nash) * win,test: change format of TEST_PIPENAME's (Santiago Gimeno) * win,pipe: fixes in uv_pipe_connect() (Santiago Gimeno) * misc: fix return value of memory functions (theanarkh) * src: add new metrics APIs (Trevor Norris) * thread: add uv_thread_getcpu() (daomingq) * build: don't use ifaddrs.h on solaris 10 (Edward Humes) * unix,win: add uv_get_available_memory() (Tim Besard) * test: fix -Wunused-but-set-variable warnings (Ben Noordhuis) * doc: bump min supported linux and freebsd versions (Ben Noordhuis) * Add Socket Runtime to the LINKS.md (Sergey Rubanov) * unix: drop kfreebsd support (Ben Noordhuis) * win: fix fstat for pipes and character files (Stefan Stojanovic) * win: fix -Wunused-variable warning (Ben Noordhuis) * win: fix -Wunused-function warning (Ben Noordhuis) * build: drop qemu-alpha from ci matrix (Ben Noordhuis) * win: move child_stdio_buffer out of uv_process_t (Santiago Gimeno) * test: fix some unreachable code warnings (Santiago Gimeno) * linux: simplify uv_uptime() (Ben Noordhuis) * test: unflake fs_event_watch_dir test (Ben Noordhuis) * darwin: remove unused fsevents symbol lookups (Ben Noordhuis) * build: add define guard around UV_EXTERN (Zvicii) * build: add UndefinedBehaviorSanitizer support (Ben Noordhuis) * build: enable platform_output test on qemu (Ben Noordhuis) * linux: handle cpu hotplugging in uv_cpu_info() (Ben Noordhuis) * build: remove unnecessary policy setting (dundargoc) * docs: add vcpkg instruction step (Jack·Boos·Yu) * win,fs: fix readlink errno for a non-symlink file (Darshan Sen) * misc: extend getpw to take uid as an argument (Jameson Nash) * unix,win: use static_assert when available (Ben Noordhuis) * docs: delete code Makefile (Jameson Nash) * docs: add CI for docs PRs (Jameson Nash) * docs: update Sphinx version on RTD (Jameson Nash) * doc: clean up license file (Ben Noordhuis) * test: fix some warnings when compiling tests (panran) * build,win: add mingw-w64 CI configuration (Jameson Nash) * build: add CI for distcheck (Jameson Nash) * unix: remove busy loop from uv_async_send (Jameson Nash) * doc: document uv_fs_cb type (Tamás Bálint Misius) * build: Improve build by cmake for Cygwin (erw7) * build: add libuv:: namespace to libuvConfig.cmake (AJ Heller) * test: fix ThreadSanitizer thread leak warning (Ben Noordhuis) * test: fix ThreadSanitizer data race warning (Ben Noordhuis) * test: fix ThreadSanitizer data race warning (Ben Noordhuis) * test: fix ThreadSanitizer data race warning (Ben Noordhuis) * test: cond-skip fork_threadpool_queue_work_simple (Ben Noordhuis) * test: cond-skip signal_multiple_loops (Ben Noordhuis) * test: cond-skip tcp_writealot (Ben Noordhuis) * build: promote tsan ci to must-pass (Ben Noordhuis) * build: add CI for OpenBSD and FreeBSD (James McCoy) * build,test: fix distcheck errors (Jameson Nash) * test: remove bad tty window size assumption (Ben Noordhuis) * darwin,process: feed kevent the signal to reap children (Jameson Nash) * unix: abort on clock_gettime() error (Ben Noordhuis) * test: remove timing-sensitive check (Ben Noordhuis) * unix: DRY and fix tcp bind error path (Jameson Nash) * macos: fix fsevents thread race conditions (Ben Noordhuis) * win: fix leak in uv_chdir (Trevor Norris) * test: make valgrind happy (Trevor Norris) * barrier: wait for prior out before next in (Jameson Nash) * test: fix visual studio 2015 build error (Ben Noordhuis) * linux: fix ceph copy error truncating readonly files (Bruno Passeri) * test: silence more valgrind warnings (Trevor Norris) * doc: add entries to LINKS.md (Trevor Norris) * win,unix: change execution order of timers (Trevor Norris) * doc: add trevnorris to maintainers (Trevor Norris) * linux: remove epoll_pwait() emulation code path (Ben Noordhuis) * linux: replace unsafe macro with inline function (Ben Noordhuis) * linux: remove arm oabi support (Ben Noordhuis) * unix,sunos: SO_REUSEPORT not valid on all sockets (Stacey Marshall) * doc: consistent single backquote in misc.rst (Jason Zhang) * src: switch to use C11 atomics where available (Trevor Norris) * test: don't use static buffer for formatting (Ben Noordhuis) * linux: introduce io_uring support (Ben Noordhuis) * linux: fix academic valgrind warning (Ben Noordhuis) * test: disable signal test under ASan and MSan (Ben Noordhuis) * linux: add IORING_OP_OPENAT support (Ben Noordhuis) * linux: add IORING_OP_CLOSE support (Ben Noordhuis) * linux: remove bug workaround for obsolete kernels (Ben Noordhuis) * doc: update active maintainers list (Ben Noordhuis) * test: add ASSERT_OK (Trevor Norris) * src: fix events/events_waiting metrics counter (Trevor Norris) * unix,win: add uv_clock_gettime() (Ben Noordhuis) * build: remove freebsd and openbsd buildbots (Ben Noordhuis) * win: fix race condition in uv__init_console() (sivadeilra) * linux: fix logic bug in sqe ring space check (Ben Noordhuis) * linux: use io_uring to batch epoll_ctl calls (Ben Noordhuis) * macos: update minimum supported version (Santiago Gimeno) * docs: fix some typos (cui fliter) * unix: use memcpy() instead of type punning (Ben Noordhuis) * test: add additional assert (Mohammed Keyvanzadeh) * build: export compile_commands.json (Lewis Russell) * win,process: write minidumps when sending SIGQUIT (Elliot Saba) * unix: constrained_memory should return UINT64_MAX (Tim Besard) * unix: handle CQ overflow in iou ring (Santiago Gimeno) * unix: remove clang compiler warning pragmas (Ben Noordhuis) * win: fix mingw build (gengjiawen) * test: fix -Wbool-compare compiler warning (Ben Noordhuis) * win: define MiniDumpWithAvxXStateContext always (Santiago Gimeno) * freebsd: hard-code UV_ENODATA definition (Santiago Gimeno) * linux: work around EOWNERDEAD io_uring kernel bug (Ben Noordhuis) * linux: fix WRITEV with lots of bufs using io_uring (Santiago Gimeno) -----BEGIN PGP SIGNATURE----- iQIzBAABCgAdFiEEYS8OrZQBYiN530QC8ow8jaM8A74FAmRnW50ACgkQ8ow8jaM8 A76noBAAiavsCBvAC554x4Naic+B3fEYjKnLe8OLTbdN4XQsS2uSkCEuP0TuwLE7 Q93CHgDFcL//XmNAXwIKMIvZ7G25zga93qxrZdd7Hsuthg5oJwsrZb8faHG7CeKy Fq+SLCUdLcwOR9m9RDp1u3h/pIhoqQvLUlPcBvDKVaJOVhIngawKU/vC125z6ZJs CYaKMlPbU2WjIkkoQjNRYL8oolmWTs5T/8AhIVWntTCyc/+/n7uxmWwlfL2iR8Kj u5IT7Vhh6a7VC+dGh26k/FN0RGdAgJCZ1Jnrt+llHxRVuw0xEBnrTAyS3hOkA7Me s+8VUaCsreVi9p66hqE5b4IL5e9yUEB18A5dQBY+TFNeXaQETVg9BxPMXbhoh+XB 2643LZn/kN1iilZBMSMl9kqB66qeUihv8cvjMnrTzq+nSxJvuc13XPfbH+zhMkVh CQaFMq3v5vc4Kyk2YrjcsfGCIkRxRGlhSmk5riz1Xf+erf4GuOl+o2fA6YvVDAnE rzlzpGoRSHgHYC23KirkBHo7qQVbWy+NeLWFRFSHfSN4+czttdusiFSjxERavfSh ezRKH6dNHI8C9fqT3oAitXOgFt6AsMiYMt7JpJwPm0SdAFjIEkgjAfhkSQADUHz7 qezLtJoCL+bcUXE8tgRRuz+J6x4FhKBZVu5uA6fV1UuyWI74DBU= =YAOh -----END PGP SIGNATURE----- Merge tag 'v1.45.0' into merge_1.45.0
2023-05-22 19:54:46 +00:00
uv_os_fd_t fd;
int r;
src: fix events/events_waiting metrics counter (#3957) The worker pool calls all callbacks locally within the queue. So the value of nevents doesn't properly reflect that case. Increase the number of events directly from the worker pool's callback to correct this. In order to properly determine if the events_waiting counter needs to be incremented, store the timeout value at the time the event provider was called.
2023-04-17 18:48:39 +00:00
char rdata;
ASSERT_OK(uv_loop_configure(uv_default_loop(), UV_METRICS_IDLE_TIME));
uv_fs_unlink(NULL, &unlink_req, "test_file", NULL);
uv_fs_req_cleanup(&unlink_req);
ASSERT_OK(uv_prepare_init(uv_default_loop(), &prepare));
ASSERT_OK(uv_prepare_start(&prepare, fs_prepare_cb));
pool_events_counter = 0;
2023.05.19, Version 1.45.0 (Stable) Changes since version 1.44.2: * win: remove stdint-msvc2008.h (Ben Noordhuis) * android: remove pthread-fixes.c (Ben Noordhuis) * build: enable MSVC_RUNTIME_LIBRARY setting (自发对称破缺) * unix: switch to c11 atomics (Ben Noordhuis) * unix: don't accept() connections in a loop (Ben Noordhuis) * win: fix off-by-1 buffer overrun in uv_exepath() (Ben Noordhuis) * build: switch ci from macos-10.15 to macos-11 (Ben Noordhuis) * win: fix thread race in uv_cwd() and uv_chdir() (Ben Noordhuis) * unix,win: remove UV_HANDLE_SHUTTING flag (Santiago Gimeno) * win: support Windows 11 in uv_os_uname() (Luan Devecchi) * unix: fix uv_getrusage() ru_maxrss reporting (Ben Noordhuis) * doc: add note about offset -1 in uv_fs_read/write (Steven Schveighoffer) * test: fix musl libc.a dlerror() test expectation (Ben Noordhuis) * kqueue: DRY file descriptor deletion logic (Ben Noordhuis) * linux: teach uv_get_constrained_memory() cgroupsv2 (Ben Noordhuis) * build: upgrade qemu-user-static package (Ben Noordhuis) * linux: move epoll.c back into linux-core.c (Ben Noordhuis) * unix: remove pre-macos 10.8 compatibility hack (Ben Noordhuis) * unix,win: fix memory leak in uv_fs_scandir() (Ben Noordhuis) * build: restore qemu download logic (Ben Noordhuis) * win: fix uv__pipe_accept memory leak (number201724) * doc: update LINKS.md (Daniel) * unix: simplify atomic op in uv_tty_reset_mode() (Ben Noordhuis) * build: add LIBUV_BUILD_SHARED cmake option (Christian Clason) * linux: remove unused or obsolete syscall wrappers (Ben Noordhuis) * linux: merge files back into single file (Ben Noordhuis) * stream: process more than one write req per loop tick (ywave620) * unix,win: give thread pool threads an 8 MB stack (Ben Noordhuis) * build: add MemorySanitizer (MSAN) support (Ben Noordhuis) * doc: add uv_poll_cb status==UV_EBADF note (jensbjorgensen) * build: support AddressSanitizer on MSVC (Jameson Nash) * win,pipe: improve method of obtaining pid for ipc (number201724) * thread: add support for affinity (daomingq) * include: map ENODATA error code (Ben Noordhuis) * build: remove bashism from autogen.sh (Santiago Gimeno) * win,tcp,udp: remove "active streams" optimization (Saúl Ibarra Corretgé) * win: drop code checking for Windows XP / Server 2k3 (Saúl Ibarra Corretgé) * unix,win: fix 'sprintf' is deprecated warning (twosee) * doc: mention close_cb can be NULL (Qix) * win: optimize udp receive performance (ywave620) * win: fix an incompatible types warning (twosee) * doc: document 0 return value for free/total memory (Ben Noordhuis) * darwin: use hw.cpufrequency again for frequency info (Jameson Nash) * win,test: change format of TEST_PIPENAME's (Santiago Gimeno) * win,pipe: fixes in uv_pipe_connect() (Santiago Gimeno) * misc: fix return value of memory functions (theanarkh) * src: add new metrics APIs (Trevor Norris) * thread: add uv_thread_getcpu() (daomingq) * build: don't use ifaddrs.h on solaris 10 (Edward Humes) * unix,win: add uv_get_available_memory() (Tim Besard) * test: fix -Wunused-but-set-variable warnings (Ben Noordhuis) * doc: bump min supported linux and freebsd versions (Ben Noordhuis) * Add Socket Runtime to the LINKS.md (Sergey Rubanov) * unix: drop kfreebsd support (Ben Noordhuis) * win: fix fstat for pipes and character files (Stefan Stojanovic) * win: fix -Wunused-variable warning (Ben Noordhuis) * win: fix -Wunused-function warning (Ben Noordhuis) * build: drop qemu-alpha from ci matrix (Ben Noordhuis) * win: move child_stdio_buffer out of uv_process_t (Santiago Gimeno) * test: fix some unreachable code warnings (Santiago Gimeno) * linux: simplify uv_uptime() (Ben Noordhuis) * test: unflake fs_event_watch_dir test (Ben Noordhuis) * darwin: remove unused fsevents symbol lookups (Ben Noordhuis) * build: add define guard around UV_EXTERN (Zvicii) * build: add UndefinedBehaviorSanitizer support (Ben Noordhuis) * build: enable platform_output test on qemu (Ben Noordhuis) * linux: handle cpu hotplugging in uv_cpu_info() (Ben Noordhuis) * build: remove unnecessary policy setting (dundargoc) * docs: add vcpkg instruction step (Jack·Boos·Yu) * win,fs: fix readlink errno for a non-symlink file (Darshan Sen) * misc: extend getpw to take uid as an argument (Jameson Nash) * unix,win: use static_assert when available (Ben Noordhuis) * docs: delete code Makefile (Jameson Nash) * docs: add CI for docs PRs (Jameson Nash) * docs: update Sphinx version on RTD (Jameson Nash) * doc: clean up license file (Ben Noordhuis) * test: fix some warnings when compiling tests (panran) * build,win: add mingw-w64 CI configuration (Jameson Nash) * build: add CI for distcheck (Jameson Nash) * unix: remove busy loop from uv_async_send (Jameson Nash) * doc: document uv_fs_cb type (Tamás Bálint Misius) * build: Improve build by cmake for Cygwin (erw7) * build: add libuv:: namespace to libuvConfig.cmake (AJ Heller) * test: fix ThreadSanitizer thread leak warning (Ben Noordhuis) * test: fix ThreadSanitizer data race warning (Ben Noordhuis) * test: fix ThreadSanitizer data race warning (Ben Noordhuis) * test: fix ThreadSanitizer data race warning (Ben Noordhuis) * test: cond-skip fork_threadpool_queue_work_simple (Ben Noordhuis) * test: cond-skip signal_multiple_loops (Ben Noordhuis) * test: cond-skip tcp_writealot (Ben Noordhuis) * build: promote tsan ci to must-pass (Ben Noordhuis) * build: add CI for OpenBSD and FreeBSD (James McCoy) * build,test: fix distcheck errors (Jameson Nash) * test: remove bad tty window size assumption (Ben Noordhuis) * darwin,process: feed kevent the signal to reap children (Jameson Nash) * unix: abort on clock_gettime() error (Ben Noordhuis) * test: remove timing-sensitive check (Ben Noordhuis) * unix: DRY and fix tcp bind error path (Jameson Nash) * macos: fix fsevents thread race conditions (Ben Noordhuis) * win: fix leak in uv_chdir (Trevor Norris) * test: make valgrind happy (Trevor Norris) * barrier: wait for prior out before next in (Jameson Nash) * test: fix visual studio 2015 build error (Ben Noordhuis) * linux: fix ceph copy error truncating readonly files (Bruno Passeri) * test: silence more valgrind warnings (Trevor Norris) * doc: add entries to LINKS.md (Trevor Norris) * win,unix: change execution order of timers (Trevor Norris) * doc: add trevnorris to maintainers (Trevor Norris) * linux: remove epoll_pwait() emulation code path (Ben Noordhuis) * linux: replace unsafe macro with inline function (Ben Noordhuis) * linux: remove arm oabi support (Ben Noordhuis) * unix,sunos: SO_REUSEPORT not valid on all sockets (Stacey Marshall) * doc: consistent single backquote in misc.rst (Jason Zhang) * src: switch to use C11 atomics where available (Trevor Norris) * test: don't use static buffer for formatting (Ben Noordhuis) * linux: introduce io_uring support (Ben Noordhuis) * linux: fix academic valgrind warning (Ben Noordhuis) * test: disable signal test under ASan and MSan (Ben Noordhuis) * linux: add IORING_OP_OPENAT support (Ben Noordhuis) * linux: add IORING_OP_CLOSE support (Ben Noordhuis) * linux: remove bug workaround for obsolete kernels (Ben Noordhuis) * doc: update active maintainers list (Ben Noordhuis) * test: add ASSERT_OK (Trevor Norris) * src: fix events/events_waiting metrics counter (Trevor Norris) * unix,win: add uv_clock_gettime() (Ben Noordhuis) * build: remove freebsd and openbsd buildbots (Ben Noordhuis) * win: fix race condition in uv__init_console() (sivadeilra) * linux: fix logic bug in sqe ring space check (Ben Noordhuis) * linux: use io_uring to batch epoll_ctl calls (Ben Noordhuis) * macos: update minimum supported version (Santiago Gimeno) * docs: fix some typos (cui fliter) * unix: use memcpy() instead of type punning (Ben Noordhuis) * test: add additional assert (Mohammed Keyvanzadeh) * build: export compile_commands.json (Lewis Russell) * win,process: write minidumps when sending SIGQUIT (Elliot Saba) * unix: constrained_memory should return UINT64_MAX (Tim Besard) * unix: handle CQ overflow in iou ring (Santiago Gimeno) * unix: remove clang compiler warning pragmas (Ben Noordhuis) * win: fix mingw build (gengjiawen) * test: fix -Wbool-compare compiler warning (Ben Noordhuis) * win: define MiniDumpWithAvxXStateContext always (Santiago Gimeno) * freebsd: hard-code UV_ENODATA definition (Santiago Gimeno) * linux: work around EOWNERDEAD io_uring kernel bug (Ben Noordhuis) * linux: fix WRITEV with lots of bufs using io_uring (Santiago Gimeno) -----BEGIN PGP SIGNATURE----- iQIzBAABCgAdFiEEYS8OrZQBYiN530QC8ow8jaM8A74FAmRnW50ACgkQ8ow8jaM8 A76noBAAiavsCBvAC554x4Naic+B3fEYjKnLe8OLTbdN4XQsS2uSkCEuP0TuwLE7 Q93CHgDFcL//XmNAXwIKMIvZ7G25zga93qxrZdd7Hsuthg5oJwsrZb8faHG7CeKy Fq+SLCUdLcwOR9m9RDp1u3h/pIhoqQvLUlPcBvDKVaJOVhIngawKU/vC125z6ZJs CYaKMlPbU2WjIkkoQjNRYL8oolmWTs5T/8AhIVWntTCyc/+/n7uxmWwlfL2iR8Kj u5IT7Vhh6a7VC+dGh26k/FN0RGdAgJCZ1Jnrt+llHxRVuw0xEBnrTAyS3hOkA7Me s+8VUaCsreVi9p66hqE5b4IL5e9yUEB18A5dQBY+TFNeXaQETVg9BxPMXbhoh+XB 2643LZn/kN1iilZBMSMl9kqB66qeUihv8cvjMnrTzq+nSxJvuc13XPfbH+zhMkVh CQaFMq3v5vc4Kyk2YrjcsfGCIkRxRGlhSmk5riz1Xf+erf4GuOl+o2fA6YvVDAnE rzlzpGoRSHgHYC23KirkBHo7qQVbWy+NeLWFRFSHfSN4+czttdusiFSjxERavfSh ezRKH6dNHI8C9fqT3oAitXOgFt6AsMiYMt7JpJwPm0SdAFjIEkgjAfhkSQADUHz7 qezLtJoCL+bcUXE8tgRRuz+J6x4FhKBZVu5uA6fV1UuyWI74DBU= =YAOh -----END PGP SIGNATURE----- Merge tag 'v1.45.0' into merge_1.45.0
2023-05-22 19:54:46 +00:00
r = uv_fs_open(NULL,
win: stop using deprecated names (#4253)
2023-12-22 11:30:48 +00:00
&open_req, "test_file", UV_FS_O_WRONLY | UV_FS_O_CREAT,
src: fix events/events_waiting metrics counter (#3957) The worker pool calls all callbacks locally within the queue. So the value of nevents doesn't properly reflect that case. Increase the number of events directly from the worker pool's callback to correct this. In order to properly determine if the events_waiting counter needs to be incremented, store the timeout value at the time the event provider was called.
2023-04-17 18:48:39 +00:00
S_IRUSR | S_IWUSR,
NULL);
2023.05.19, Version 1.45.0 (Stable) Changes since version 1.44.2: * win: remove stdint-msvc2008.h (Ben Noordhuis) * android: remove pthread-fixes.c (Ben Noordhuis) * build: enable MSVC_RUNTIME_LIBRARY setting (自发对称破缺) * unix: switch to c11 atomics (Ben Noordhuis) * unix: don't accept() connections in a loop (Ben Noordhuis) * win: fix off-by-1 buffer overrun in uv_exepath() (Ben Noordhuis) * build: switch ci from macos-10.15 to macos-11 (Ben Noordhuis) * win: fix thread race in uv_cwd() and uv_chdir() (Ben Noordhuis) * unix,win: remove UV_HANDLE_SHUTTING flag (Santiago Gimeno) * win: support Windows 11 in uv_os_uname() (Luan Devecchi) * unix: fix uv_getrusage() ru_maxrss reporting (Ben Noordhuis) * doc: add note about offset -1 in uv_fs_read/write (Steven Schveighoffer) * test: fix musl libc.a dlerror() test expectation (Ben Noordhuis) * kqueue: DRY file descriptor deletion logic (Ben Noordhuis) * linux: teach uv_get_constrained_memory() cgroupsv2 (Ben Noordhuis) * build: upgrade qemu-user-static package (Ben Noordhuis) * linux: move epoll.c back into linux-core.c (Ben Noordhuis) * unix: remove pre-macos 10.8 compatibility hack (Ben Noordhuis) * unix,win: fix memory leak in uv_fs_scandir() (Ben Noordhuis) * build: restore qemu download logic (Ben Noordhuis) * win: fix uv__pipe_accept memory leak (number201724) * doc: update LINKS.md (Daniel) * unix: simplify atomic op in uv_tty_reset_mode() (Ben Noordhuis) * build: add LIBUV_BUILD_SHARED cmake option (Christian Clason) * linux: remove unused or obsolete syscall wrappers (Ben Noordhuis) * linux: merge files back into single file (Ben Noordhuis) * stream: process more than one write req per loop tick (ywave620) * unix,win: give thread pool threads an 8 MB stack (Ben Noordhuis) * build: add MemorySanitizer (MSAN) support (Ben Noordhuis) * doc: add uv_poll_cb status==UV_EBADF note (jensbjorgensen) * build: support AddressSanitizer on MSVC (Jameson Nash) * win,pipe: improve method of obtaining pid for ipc (number201724) * thread: add support for affinity (daomingq) * include: map ENODATA error code (Ben Noordhuis) * build: remove bashism from autogen.sh (Santiago Gimeno) * win,tcp,udp: remove "active streams" optimization (Saúl Ibarra Corretgé) * win: drop code checking for Windows XP / Server 2k3 (Saúl Ibarra Corretgé) * unix,win: fix 'sprintf' is deprecated warning (twosee) * doc: mention close_cb can be NULL (Qix) * win: optimize udp receive performance (ywave620) * win: fix an incompatible types warning (twosee) * doc: document 0 return value for free/total memory (Ben Noordhuis) * darwin: use hw.cpufrequency again for frequency info (Jameson Nash) * win,test: change format of TEST_PIPENAME's (Santiago Gimeno) * win,pipe: fixes in uv_pipe_connect() (Santiago Gimeno) * misc: fix return value of memory functions (theanarkh) * src: add new metrics APIs (Trevor Norris) * thread: add uv_thread_getcpu() (daomingq) * build: don't use ifaddrs.h on solaris 10 (Edward Humes) * unix,win: add uv_get_available_memory() (Tim Besard) * test: fix -Wunused-but-set-variable warnings (Ben Noordhuis) * doc: bump min supported linux and freebsd versions (Ben Noordhuis) * Add Socket Runtime to the LINKS.md (Sergey Rubanov) * unix: drop kfreebsd support (Ben Noordhuis) * win: fix fstat for pipes and character files (Stefan Stojanovic) * win: fix -Wunused-variable warning (Ben Noordhuis) * win: fix -Wunused-function warning (Ben Noordhuis) * build: drop qemu-alpha from ci matrix (Ben Noordhuis) * win: move child_stdio_buffer out of uv_process_t (Santiago Gimeno) * test: fix some unreachable code warnings (Santiago Gimeno) * linux: simplify uv_uptime() (Ben Noordhuis) * test: unflake fs_event_watch_dir test (Ben Noordhuis) * darwin: remove unused fsevents symbol lookups (Ben Noordhuis) * build: add define guard around UV_EXTERN (Zvicii) * build: add UndefinedBehaviorSanitizer support (Ben Noordhuis) * build: enable platform_output test on qemu (Ben Noordhuis) * linux: handle cpu hotplugging in uv_cpu_info() (Ben Noordhuis) * build: remove unnecessary policy setting (dundargoc) * docs: add vcpkg instruction step (Jack·Boos·Yu) * win,fs: fix readlink errno for a non-symlink file (Darshan Sen) * misc: extend getpw to take uid as an argument (Jameson Nash) * unix,win: use static_assert when available (Ben Noordhuis) * docs: delete code Makefile (Jameson Nash) * docs: add CI for docs PRs (Jameson Nash) * docs: update Sphinx version on RTD (Jameson Nash) * doc: clean up license file (Ben Noordhuis) * test: fix some warnings when compiling tests (panran) * build,win: add mingw-w64 CI configuration (Jameson Nash) * build: add CI for distcheck (Jameson Nash) * unix: remove busy loop from uv_async_send (Jameson Nash) * doc: document uv_fs_cb type (Tamás Bálint Misius) * build: Improve build by cmake for Cygwin (erw7) * build: add libuv:: namespace to libuvConfig.cmake (AJ Heller) * test: fix ThreadSanitizer thread leak warning (Ben Noordhuis) * test: fix ThreadSanitizer data race warning (Ben Noordhuis) * test: fix ThreadSanitizer data race warning (Ben Noordhuis) * test: fix ThreadSanitizer data race warning (Ben Noordhuis) * test: cond-skip fork_threadpool_queue_work_simple (Ben Noordhuis) * test: cond-skip signal_multiple_loops (Ben Noordhuis) * test: cond-skip tcp_writealot (Ben Noordhuis) * build: promote tsan ci to must-pass (Ben Noordhuis) * build: add CI for OpenBSD and FreeBSD (James McCoy) * build,test: fix distcheck errors (Jameson Nash) * test: remove bad tty window size assumption (Ben Noordhuis) * darwin,process: feed kevent the signal to reap children (Jameson Nash) * unix: abort on clock_gettime() error (Ben Noordhuis) * test: remove timing-sensitive check (Ben Noordhuis) * unix: DRY and fix tcp bind error path (Jameson Nash) * macos: fix fsevents thread race conditions (Ben Noordhuis) * win: fix leak in uv_chdir (Trevor Norris) * test: make valgrind happy (Trevor Norris) * barrier: wait for prior out before next in (Jameson Nash) * test: fix visual studio 2015 build error (Ben Noordhuis) * linux: fix ceph copy error truncating readonly files (Bruno Passeri) * test: silence more valgrind warnings (Trevor Norris) * doc: add entries to LINKS.md (Trevor Norris) * win,unix: change execution order of timers (Trevor Norris) * doc: add trevnorris to maintainers (Trevor Norris) * linux: remove epoll_pwait() emulation code path (Ben Noordhuis) * linux: replace unsafe macro with inline function (Ben Noordhuis) * linux: remove arm oabi support (Ben Noordhuis) * unix,sunos: SO_REUSEPORT not valid on all sockets (Stacey Marshall) * doc: consistent single backquote in misc.rst (Jason Zhang) * src: switch to use C11 atomics where available (Trevor Norris) * test: don't use static buffer for formatting (Ben Noordhuis) * linux: introduce io_uring support (Ben Noordhuis) * linux: fix academic valgrind warning (Ben Noordhuis) * test: disable signal test under ASan and MSan (Ben Noordhuis) * linux: add IORING_OP_OPENAT support (Ben Noordhuis) * linux: add IORING_OP_CLOSE support (Ben Noordhuis) * linux: remove bug workaround for obsolete kernels (Ben Noordhuis) * doc: update active maintainers list (Ben Noordhuis) * test: add ASSERT_OK (Trevor Norris) * src: fix events/events_waiting metrics counter (Trevor Norris) * unix,win: add uv_clock_gettime() (Ben Noordhuis) * build: remove freebsd and openbsd buildbots (Ben Noordhuis) * win: fix race condition in uv__init_console() (sivadeilra) * linux: fix logic bug in sqe ring space check (Ben Noordhuis) * linux: use io_uring to batch epoll_ctl calls (Ben Noordhuis) * macos: update minimum supported version (Santiago Gimeno) * docs: fix some typos (cui fliter) * unix: use memcpy() instead of type punning (Ben Noordhuis) * test: add additional assert (Mohammed Keyvanzadeh) * build: export compile_commands.json (Lewis Russell) * win,process: write minidumps when sending SIGQUIT (Elliot Saba) * unix: constrained_memory should return UINT64_MAX (Tim Besard) * unix: handle CQ overflow in iou ring (Santiago Gimeno) * unix: remove clang compiler warning pragmas (Ben Noordhuis) * win: fix mingw build (gengjiawen) * test: fix -Wbool-compare compiler warning (Ben Noordhuis) * win: define MiniDumpWithAvxXStateContext always (Santiago Gimeno) * freebsd: hard-code UV_ENODATA definition (Santiago Gimeno) * linux: work around EOWNERDEAD io_uring kernel bug (Ben Noordhuis) * linux: fix WRITEV with lots of bufs using io_uring (Santiago Gimeno) -----BEGIN PGP SIGNATURE----- iQIzBAABCgAdFiEEYS8OrZQBYiN530QC8ow8jaM8A74FAmRnW50ACgkQ8ow8jaM8 A76noBAAiavsCBvAC554x4Naic+B3fEYjKnLe8OLTbdN4XQsS2uSkCEuP0TuwLE7 Q93CHgDFcL//XmNAXwIKMIvZ7G25zga93qxrZdd7Hsuthg5oJwsrZb8faHG7CeKy Fq+SLCUdLcwOR9m9RDp1u3h/pIhoqQvLUlPcBvDKVaJOVhIngawKU/vC125z6ZJs CYaKMlPbU2WjIkkoQjNRYL8oolmWTs5T/8AhIVWntTCyc/+/n7uxmWwlfL2iR8Kj u5IT7Vhh6a7VC+dGh26k/FN0RGdAgJCZ1Jnrt+llHxRVuw0xEBnrTAyS3hOkA7Me s+8VUaCsreVi9p66hqE5b4IL5e9yUEB18A5dQBY+TFNeXaQETVg9BxPMXbhoh+XB 2643LZn/kN1iilZBMSMl9kqB66qeUihv8cvjMnrTzq+nSxJvuc13XPfbH+zhMkVh CQaFMq3v5vc4Kyk2YrjcsfGCIkRxRGlhSmk5riz1Xf+erf4GuOl+o2fA6YvVDAnE rzlzpGoRSHgHYC23KirkBHo7qQVbWy+NeLWFRFSHfSN4+czttdusiFSjxERavfSh ezRKH6dNHI8C9fqT3oAitXOgFt6AsMiYMt7JpJwPm0SdAFjIEkgjAfhkSQADUHz7 qezLtJoCL+bcUXE8tgRRuz+J6x4FhKBZVu5uA6fV1UuyWI74DBU= =YAOh -----END PGP SIGNATURE----- Merge tag 'v1.45.0' into merge_1.45.0
2023-05-22 19:54:46 +00:00
ASSERT_EQ(r, 0);
ASSERT_GE(open_req.result, 0);
fd = (uv_os_fd_t) open_req.result;
src: fix events/events_waiting metrics counter (#3957) The worker pool calls all callbacks locally within the queue. So the value of nevents doesn't properly reflect that case. Increase the number of events directly from the worker pool's callback to correct this. In order to properly determine if the events_waiting counter needs to be incremented, store the timeout value at the time the event provider was called.
2023-04-17 18:48:39 +00:00
uv_fs_req_cleanup(&open_req);
iov = uv_buf_init(test_buf, sizeof(test_buf));
ASSERT_OK(uv_fs_write(uv_default_loop(),
&write_req,
fd,
&iov,
1,
0,
fs_write_cb));
ASSERT_OK(uv_fs_stat(uv_default_loop(),
&stat1_req,
"test_file",
fs_stat_cb));
ASSERT_OK(uv_fs_stat(uv_default_loop(),
&stat2_req,
"test_file",
fs_stat_cb));
ASSERT_OK(uv_random(uv_default_loop(),
&random_req,
&rdata,
1,
0,
fs_random_cb));
ASSERT_OK(uv_getaddrinfo(uv_default_loop(),
&addrinfo_req,
fs_addrinfo_cb,
"example.invalid",
NULL,
NULL));
/* Sleep for a moment to hopefully force the events to complete before
* entering the event loop. */
uv_sleep(100);
ASSERT_OK(uv_run(uv_default_loop(), UV_RUN_DEFAULT));
ASSERT_OK(uv_metrics_info(uv_default_loop(), &metrics));
/* It's possible for uv__work_done() to execute one extra time even though the
* QUEUE has already been cleared out. This has to do with the way we use an
* uv_async to tell the event loop thread to process the worker pool QUEUE. */
ASSERT_GE(metrics.events, 7);
/* It's possible one of the other events also got stuck in the event queue, so
* check GE instead of EQ. Reason for 4 instead of 5 is because the call to
* uv_getaddrinfo() is racey and slow. So can't guarantee that it'll always
* execute before sleep completes. */
ASSERT_GE(metrics.events_waiting, 4);
ASSERT_EQ(pool_events_counter, -42);
uv_fs_unlink(NULL, &unlink_req, "test_file", NULL);
uv_fs_req_cleanup(&unlink_req);
MAKE_VALGRIND_HAPPY(uv_default_loop());
core: add API to measure event loop idle time The API addition `uv_metrics_idle_time()` is a thread safe call that allows the user to retrieve the amount of time the event loop has spent in the kernel's event provider (i.e. poll). It was done this way to allow retrieving this value without needing to interrupt the execution of the event loop. This option can be enabled by passing `UV_METRICS_IDLE_TIME` to `uv_loop_configure()`. One important aspect of this change is, when enabled, to always first call the event provider with a `timeout == 0`. This allows libuv to know whether any events were waiting in the event queue when the event provider was called. The importance of this is because libuv is tracking the amount of "idle time", not "poll time". Thus the provider entry time is not recorded when `timeout == 0` (the event provider never idles in this case). While this does add a small amount of overhead, when enabled, but the overhead decreases when the event loop has a heavier load. This is because poll events will be waiting when the event provider is called. Thus never actually recording the provider entry time. Checking if `uv_loop_t` is configured with `UV_METRICS_IDLE_TIME` always happens in `uv__metrics_set_provider_entry_time()` and `uv__metrics_update_idle_time()`. Making the conditional logic wrapping each call simpler and allows for instrumentation to always hook into those two function calls. Rather than placing the fields directly on `uv__loop_internal_fields_t` add the struct `uv__loop_metrics_t` as a location for future metrics API additions. Tests and additional documentation has been included. PR-URL: https://github.com/libuv/libuv/pull/2725 Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com> Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Jameson Nash <vtjnash@gmail.com>
2020-03-26 23:15:13 +00:00
return 0;
}
Reference in New Issue Copy Permalink