From: Stefan Eilemann <eilemann@gmail.com>
To: Pthreads-Win32 list <pthreads-win32@sourceware.org>
Subject: pthread_exit and STL problems
Date: Tue, 30 Jan 2007 09:20:00 -0000 [thread overview]
Message-ID: <3D7E7834-79B3-4C11-BA29-DBF2821B8B6B@gmail.com> (raw)
Hi,
I am tracking down a problem in my code using VC++ 8.0 and
pthreads-win32 2.8.0. My hope is that somebody on the list has
seen (and solved) this problem.
In my code I am removing elements from a std::vector<>, which
works fine until the first pthread has exited. After this point,
I am getting a crash in the debug version when erasing an element
from the vector:
vector<Object*>::iterator iter = find( objects.begin(),
objects.end(),
object );
if( iter == objects.end( ))
return;
objects.erase( iter );
This is the call stack:
Equalizer.dll!std::vector<eqNet::Object
*,std::allocator<eqNet::Object *> >::_Orphan_range(eqNet::Object * *
_First=0x0038c830, eqNet::Object * * _Last=0x0038c834) Line 1233 +
0x5 bytes
Equalizer.dll!std::vector<eqNet::Object
*,std::allocator<eqNet::Object *> >::erase
(std::_Vector_iterator<eqNet::Object *,std::allocator<eqNet::Object
*> > _Where=0x00395fe0 {_initData={...} _frameData={...}
_displayLists={...} }) Line 990
Equalizer.dll!eqNet::Session::removeRegisteredObject(eqNet::Object *
object=0x00395fe0, eqNet::Object::SharePolicy policy=SHARE_NODE)
Line 304 + 0x29 bytes
Equalizer.dll!eq::Node::_cmdDestroyPipe(eqNet::Command & command=
{...}) Line 134 C++
The code in question seems to invalidate all iterators of the vector.
The crash happens because _Pnext gets an invalid pointer after some
iterations:
#if _HAS_ITERATOR_DEBUGGING
void _Orphan_range(pointer _First, pointer _Last) const
{ // orphan iterators within specified (inclusive) range
_Lockit _Lock(_LOCK_DEBUG);
const_iterator **_Pnext = (const_iterator **)&this->_Myfirstiter;
while (*_Pnext != 0)
if ((*_Pnext)->_Myptr < _First || _Last < (*_Pnext)->_Myptr)
_Pnext = (const_iterator **)&(*_Pnext)->_Mynextiter;
else
{ // orphan the iterator
(*_Pnext)->_Mycont = 0;
*_Pnext = (const_iterator *)(*_Pnext)->_Mynextiter;
}
}
#endif /* _HAS_ITERATOR_DEBUGGING */
Again - this code is called numerous time before without problem. The
first time the crash happens is when a pthread has been exit'ed, and
the calling thread just called pthread_join() shortly before the call
above. I am linking against the multithreaded debug runtime.
Has anybody seen this already? Does the thread exit code cleanup/
interfere
with the VC8 STL implementation? Any hints?
Cheers,
Stefan.
--
http://www.equalizergraphics.com
http://www.linkedin.com/in/eilemann
next reply other threads:[~2007-01-30 9:20 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-01-30 9:20 Stefan Eilemann [this message]
[not found] ` <45BFD127.6040609@homemail.com.au>
2007-01-31 9:02 ` Stefan Eilemann
2007-01-31 13:54 ` Streithorst, Kip
[not found] ` <45C0A0B5.6040306@homemail.com.au>
[not found] ` <A75AD195-4E38-4A02-8979-A3EC3C568FB2@gmail.com>
[not found] ` <45C1DFB1.6010508@homemail.com.au>
2007-02-02 9:11 ` Stefan Eilemann
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=3D7E7834-79B3-4C11-BA29-DBF2821B8B6B@gmail.com \
--to=eilemann@gmail.com \
--cc=pthreads-win32@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).