From strizhov at netsec.colostate.edu  Fri Jan 14 00:09:17 2011
From: strizhov at netsec.colostate.edu (Mikhail Strizhov)
Date: Thu, 13 Jan 2011 16:09:17 -0700
Subject: [Expat-discuss] Expat lib and Pthreads
Message-ID: <4D2F861D.9010209@netsec.colostate.edu>

Greetings list,

I have fully working parsing module and I need to use this module with 
two threads (C pthread library) to parse different xml documents 
simultaneously.
Both threads will launch:

XML_Parser parser = XML_ParserCreate(NULL);
XML_SetElementHandler(parser, _start_hndl, _end_hndl);
XML_SetCharacterDataHandler(parser,_char_hndl );

Will I have inconsistency if both threads will use _start_hndl() and 
_end_hndl() handler functions?

Thanks!

-- 
Sincerely,
Mikhail Strizhov
Email: strizhov at netsec.colostate.edu <mailto:strizhov at netsec.colostate.edu>

From nickmacd at gmail.com  Sat Jan 15 17:18:20 2011
From: nickmacd at gmail.com (Nick MacDonald)
Date: Sat, 15 Jan 2011 11:18:20 -0500
Subject: [Expat-discuss] Expat lib and Pthreads
In-Reply-To: <4D2F861D.9010209@netsec.colostate.edu>
References: <4D2F861D.9010209@netsec.colostate.edu>
Message-ID: <AANLkTikz8k9DtP+tDLsH9WBq7cx+pnuTZjn6DBJaSRxL@mail.gmail.com>

Potentially, yes... depending on how you intend to keep the state of
the two threads separate.  You may want to learn about the value of
User Data, and set your thread state in some data structure, the
pointer of which you would pass into eXpat with XML_SetUserData such
that it can pass it along to the handler routines and you can use it
in the handlers to get access to your thread specific state.

XML_SetUserData(XML_Parser p, void *userData)
This sets the user data pointer that gets passed to handlers.

Good luck with your project...
  Nick


On Thu, Jan 13, 2011 at 6:09 PM, Mikhail Strizhov
<strizhov at netsec.colostate.edu> wrote:
> I have fully working parsing module and I need to use this module with two
> threads (C pthread library) to parse different xml documents simultaneously.
> Both threads will launch:
>
> XML_Parser parser = XML_ParserCreate(NULL);
> XML_SetElementHandler(parser, _start_hndl, _end_hndl);
> XML_SetCharacterDataHandler(parser,_char_hndl );
>
> Will I have inconsistency if both threads will use _start_hndl() and
> _end_hndl() handler functions?

From CarstenFuchs at T-Online.de  Mon Jan 24 11:34:14 2011
From: CarstenFuchs at T-Online.de (Carsten Fuchs)
Date: Mon, 24 Jan 2011 11:34:14 +0100
Subject: [Expat-discuss] Crash in lookup()
Message-ID: <4D3D55A6.6080208@T-Online.de>

Dear expat list,

we're using expat 2.0.1 (with OpenCollada) for loading Collada documents, using Win7 with VC++ 2010.
For testing, we also tried the current CVS tarball, with the same results.

Unfortunately, when we try to load a simple sample file, we get a segfault in function
NAMED * lookup(HASH_TABLE *table, KEY name, size_t createSize)


The stacktrace without function parameters is at:
http://pastebin.com/EN9xkVnr

The same stacktrace with function parameters is at:
http://pastebin.com/EcXkkXm5

A link to the relevant line of source code is
http://trac.cafu.de/browser/cafu/trunk/ExtLibs/expat/lib/xmlparse.c#L5880

where i == 0x38 and table->v[i] == 0xdddddddd (pointer to previously deleted memory).


Is this a known bug?
What can we do?

Any help would much be appreciated!

Many thanks and best regards,
Carsten



-- 
    Cafu - the open-source Game and Graphics Engine
for multiplayer, cross-platform, real-time 3D Action
           Learn more at http://www.cafu.de

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 5219 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://mail.libexpat.org/pipermail/expat-discuss/attachments/20110124/dc5f1b5b/attachment.bin>

From CarstenFuchs at T-Online.de  Mon Jan 24 13:27:40 2011
From: CarstenFuchs at T-Online.de (Carsten Fuchs)
Date: Mon, 24 Jan 2011 13:27:40 +0100
Subject: [Expat-discuss] Crash in lookup()
In-Reply-To: <4D3D55A6.6080208@T-Online.de>
References: <4D3D55A6.6080208@T-Online.de>
Message-ID: <4D3D703C.7060609@T-Online.de>

Am 2011-01-24 11:34, schrieb Carsten Fuchs:
> where i == 0x38 and table->v[i] == 0xdddddddd (pointer to previously deleted memory).

Small correction:
table->v[i] itself is not 0xdddddddd of course, but the memory it points to is filled by C 
runtime with 0xdd bytes when it was deleted, and thus table->v[i]->name == 0xdddddddd.

Best regards,
Carsten



-- 
    Cafu - the open-source Game and Graphics Engine
for multiplayer, cross-platform, real-time 3D Action
           Learn more at http://www.cafu.de

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 5219 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://mail.libexpat.org/pipermail/expat-discuss/attachments/20110124/c4e27492/attachment.bin>