public inbox for libc-hacker@sourceware.org
 help / color / mirror / Atom feed
* [PATCH] Fix 2 NPTL testcases
@ 2006-09-18 10:48 Jakub Jelinek
  2006-09-18 13:42 ` Ulrich Drepper
  0 siblings, 1 reply; 2+ messages in thread
From: Jakub Jelinek @ 2006-09-18 10:48 UTC (permalink / raw)
  To: Ulrich Drepper; +Cc: Glibc hackers

Hi!

tst-kill4 does something apps generally shouldn't do - it relies on the
thread descriptor to be put into stack cache and not being immediately
freed, unfortunately if the stack size is larger than the limit for
stack cache, it is freed immediately and the stack size is determined
from current ulimit -s.  This test kept crashing on ia64 for a while,
but can be reproduced e.g. on ppc32 with ulimit -s 80000.

On IA-64 with its larger page size, writing 100000 bytes into a pipe
succeeds in contemporary kernels, even when nobody reads anything
from the other pipe's end.  This patch changes it to loop, so it will
eventually block.

2006-09-18  Jakub Jelinek  <jakub@redhat.com>

	* tst-kill4.c (do_test): Explicitly set tf thread's stack size.
	* tst-cancel2.c (tf): Loop as long as something was written.

--- libc/nptl/tst-kill4.c.jj	2003-02-21 21:27:54.000000000 +0100
+++ libc/nptl/tst-kill4.c	2006-09-18 12:32:48.000000000 +0200
@@ -1,4 +1,4 @@
-/* Copyright (C) 2003 Free Software Foundation, Inc.
+/* Copyright (C) 2003, 2006 Free Software Foundation, Inc.
    This file is part of the GNU C Library.
    Contributed by Ulrich Drepper <drepper@redhat.com>, 2003.
 
@@ -35,13 +35,30 @@ tf (void *a)
 int
 do_test (void)
 {
+  pthread_attr_t at;
+  if (pthread_attr_init (&at) != 0)
+    {
+      puts ("attr_create failed");
+      exit (1);
+    }
+
+  /* Limit thread stack size, because if it is too large, pthread_join
+     will free it immediately rather than put it into stack cache.  */
+  if (pthread_attr_setstacksize (&at, 2 * 1024 * 1024) != 0)
+    {
+      puts ("setstacksize failed");
+      exit (1);
+    }
+
   pthread_t th;
-  if (pthread_create (&th, NULL, tf, NULL) != 0)
+  if (pthread_create (&th, &at, tf, NULL) != 0)
     {
       puts ("create failed");
       exit (1);
     }
 
+  pthread_attr_destroy (&at);
+
   if (pthread_join (th, NULL) != 0)
     {
       puts ("join failed");
--- libc/nptl/tst-cancel2.c.jj	2002-11-26 23:49:53.000000000 +0100
+++ libc/nptl/tst-cancel2.c	2006-09-18 12:33:53.000000000 +0200
@@ -1,4 +1,4 @@
-/* Copyright (C) 2002 Free Software Foundation, Inc.
+/* Copyright (C) 2002, 2006 Free Software Foundation, Inc.
    This file is part of the GNU C Library.
    Contributed by Ulrich Drepper <drepper@redhat.com>, 2002.
 
@@ -33,11 +33,7 @@ tf (void *arg)
      write blocks.  */
   char buf[100000];
 
-  if (write (fd[1], buf, sizeof (buf)) == sizeof (buf))
-    {
-      puts ("write succeeded");
-      return (void *) 1l;
-    }
+  while (write (fd[1], buf, sizeof (buf)) > 0);
 
   return (void *) 42l;
 }

	Jakub

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH] Fix 2 NPTL testcases
  2006-09-18 10:48 [PATCH] Fix 2 NPTL testcases Jakub Jelinek
@ 2006-09-18 13:42 ` Ulrich Drepper
  0 siblings, 0 replies; 2+ messages in thread
From: Ulrich Drepper @ 2006-09-18 13:42 UTC (permalink / raw)
  To: Jakub Jelinek; +Cc: Glibc hackers

[-- Attachment #1: Type: text/plain, Size: 101 bytes --]

Applied.

-- 
➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 251 bytes --]

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2006-09-18 13:42 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2006-09-18 10:48 [PATCH] Fix 2 NPTL testcases Jakub Jelinek
2006-09-18 13:42 ` Ulrich Drepper

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).