Message-ID: <4D7FB133.7030301@dronecode.org.uk>
Date: Tue, 15 Mar 2011 18:34:27 +0000
From: Jon TURNEY <jon.turney@dronecode.org.uk>
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.2.15) Gecko/20110303 Thunderbird/3.1.9
MIME-Version: 1.0
To: cygwin@cygwin.com
Subject: Calling sem_wait() in DllMain( DLL_THREAD_DETACH) crashes
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Mailing-List: contact cygwin-help@cygwin.com; run by ezmlm
Sender: cygwin-owner@cygwin.com
Mail-Followup-To: cygwin@cygwin.com


I've come across a problem whilst trying to run the twisted test suite, where
some tests just cause python to segfault.  It seems to be the same issue with
libcrypto as reported in [2],[3]

A small test case is attached to [2] as cygwin_crash.zip, referred to in
msg76086, which shows that manipulating a semaphore in
DllMain(DLL_THREAD_DETACH) causes a crash.

There is a patch for openssl attached to [1] which makes sense to me.  Surely
the cleanup ERR_remove_state() does should be requested using
pthread_cleanup_push() rather using DllMain(DLL_THREAD_DETACH)?  But that
doesn't seem to be done for any other POSIX target.  The documentation of
ERR_remove_state() gives me the impression that it expects the application to
arrange for that to happen, so perhaps it's just trying too hard to be
helpful here?

I suppose the alternative interpretation is that you are supposed to be able
to call anything you like from DllMain(DLL_THREAD_DETACH), then this is a
cygwin bug. But I can't see how that could be made work as pthread::exit() has
to delete the cygwin thread object before ExitThread() is called, as there's
no thread to do it in afterwards :-)

[1] http://www.mail-archive.com/openssl-dev@openssl.org/msg24599.html
[2] http://bugs.python.org/issue3947
[3] http://cygwin.com/ml/cygwin/2008-11/msg00341.html

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple