public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
From: malteskarupke@fastmail.fm
To: libc-alpha@sourceware.org
Cc: Malte Skarupke <malteskarupke@fastmail.fm>
Subject: [PATCH v3 4/6] nptl: Make test-cond-printers check the number of waiters
Date: Thu,  6 Oct 2022 17:43:27 -0400	[thread overview]
Message-ID: <20221006214329.1084244-5-malteskarupke@fastmail.fm> (raw)
In-Reply-To: <20221006214329.1084244-1-malteskarupke@fastmail.fm>

From: Malte Skarupke <malteskarupke@fastmail.fm>

In my last change I changed the semantics of how to determine the
number of waiters on a condition variable. The existing test only
tested that the printers print something. They didn't cover the case
when there is a thread sleeping on the condition variable. In this
patch I changed the test to ensure that the correct number is printed.
---
 nptl/test-cond-printers.c  | 56 +++++++++++++++++++++++++++++++++-----
 nptl/test-cond-printers.py |  5 ++++
 2 files changed, 54 insertions(+), 7 deletions(-)

diff --git a/nptl/test-cond-printers.c b/nptl/test-cond-printers.c
index 51d7f920b3..615fd5a681 100644
--- a/nptl/test-cond-printers.c
+++ b/nptl/test-cond-printers.c
@@ -26,7 +26,14 @@
 #define PASS 0
 #define FAIL 1
 
-static int test_status_destroyed (pthread_cond_t *condvar);
+static int test_status (pthread_cond_t *condvar);
+
+typedef struct
+{
+  pthread_mutex_t *mutex;
+  pthread_cond_t *condvar;
+  int *wait_thread_asleep;
+} test_state;
 
 int
 main (void)
@@ -36,22 +43,57 @@ main (void)
   int result = FAIL;
 
   if (pthread_condattr_init (&attr) == 0
-      && test_status_destroyed (&condvar) == PASS)
+      && test_status (&condvar) == PASS)
     result = PASS;
   /* Else, one of the pthread_cond* functions failed.  */
 
   return result;
 }
 
+static void *
+wait (void *arg)
+{
+  test_state *state = (test_state *)arg;
+  void *result = PASS;
+  if (pthread_mutex_lock (state->mutex) != 0)
+    result = (void *)FAIL;
+  *state->wait_thread_asleep = 1;
+  if (pthread_cond_signal (state->condvar) != 0)
+    result = (void *)FAIL;
+  if (pthread_cond_wait (state->condvar, state->mutex) != 0)
+    result = (void *)FAIL;
+  if (pthread_mutex_unlock (state->mutex) != 0)
+    result = (void *)FAIL;
+  return result;
+}
+
 /* Initializes CONDVAR, then destroys it.  */
 static int
-test_status_destroyed (pthread_cond_t *condvar)
+test_status (pthread_cond_t *condvar)
 {
-  int result = FAIL;
+  int result = PASS;
 
-  if (pthread_cond_init (condvar, NULL) == 0
-      && pthread_cond_destroy (condvar) == 0)
-    result = PASS; /* Test status (destroyed).  */
+  pthread_mutex_t mutex;
+  result |= pthread_mutex_init (&mutex, NULL);
+  result |= pthread_cond_init (condvar, NULL);
+  int wait_thread_asleep = 0;
+  test_state state = { &mutex, condvar, &wait_thread_asleep };
+  result |= pthread_mutex_lock (&mutex);
+  pthread_t thread;
+  result |= pthread_create (&thread, NULL, wait, &state);
+  while (!wait_thread_asleep)
+    {
+      result |= pthread_cond_wait (condvar, &mutex);
+    }
+  result |= pthread_cond_signal (condvar); /* Test about to signal */
+  result |= pthread_mutex_unlock (&mutex);
+  result |= pthread_cond_destroy (condvar);
+  void *retval = NULL;
+  result |= pthread_join (thread, &retval);  /* Test status (destroyed).  */
+  result |= pthread_mutex_destroy (&mutex);
+  result = result ? FAIL : PASS;
+  if (retval != NULL)
+    result = FAIL;
 
   return result;
 }
diff --git a/nptl/test-cond-printers.py b/nptl/test-cond-printers.py
index 42329c1691..7945c7a0d5 100644
--- a/nptl/test-cond-printers.py
+++ b/nptl/test-cond-printers.py
@@ -33,6 +33,11 @@ try:
     var = 'condvar'
     to_string = 'pthread_cond_t'
 
+    break_at(test_source, 'Test about to signal')
+    continue_cmd() # Go to test_status_destroyed
+    test_printer(var, to_string, {'Threads known to still execute a wait function': '1'})
+
+
     break_at(test_source, 'Test status (destroyed)')
     continue_cmd() # Go to test_status_destroyed
     test_printer(var, to_string, {'Threads known to still execute a wait function': '0'})
-- 
2.25.1


  parent reply	other threads:[~2022-10-06 21:43 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-10-06 21:43 [PATCH v3 0/6] nptl: Fix pthread_cond_signal missing a sleeper malteskarupke
2022-10-06 21:43 ` [PATCH v3 1/6] nptl: Fix pthread_cond_signal missing a sleeper (#BZ 25847) malteskarupke
2022-10-06 21:43 ` [PATCH v3 2/6] nptl: Remove the signal-stealing code. It is no longer needed malteskarupke
2022-10-06 21:43 ` [PATCH v3 3/6] nptl: Optimization by not incrementing wrefs in pthread_cond_wait malteskarupke
2022-10-06 21:43 ` malteskarupke [this message]
2022-10-06 21:43 ` [PATCH v3 5/6] nptl: Rename __wrefs to __crefs because its meaning has changed malteskarupke
2022-10-06 21:43 ` [PATCH v3 6/6] nptl: Cleaning up __g1_start and related code in pthread_cond_wait malteskarupke

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20221006214329.1084244-5-malteskarupke@fastmail.fm \
    --to=malteskarupke@fastmail.fm \
    --cc=libc-alpha@sourceware.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).