From: Andris Pavenis <pavenis AT latnet DOT lv>
To: djgpp AT delorie DOT com, eplmst AT lu DOT erisoft DOT se (Martin Stromberg)
Subject: Re: SIGILL 386 (illegal opcode)
Date: Fri, 16 May 2003 08:14:46 +0300
User-Agent: KMail/1.5.1
References: <b8oqgi$1s2b$1 AT hq DOT mark-itt DOT ru> <b9tirj$q54$1 AT hq DOT mark-itt DOT ru> <b9tpv7$j93$1 AT antares DOT lu DOT erisoft DOT se>
In-Reply-To: <b9tpv7$j93$1@antares.lu.erisoft.se>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200305160814.46240.pavenis@latnet.lv>
Reply-To: djgpp AT delorie DOT com

On Wednesday 14 May 2003 19:13, Martin Stromberg wrote:
> Alexei K. (kzua AT kzua DOT udm DOT ru) wrote:
> : Thanks a lot. As I see it is not problem of DJGPP libstdcxx.
> : I found in libstdc++-v3 FAQ the following article:
> :
> : "Threading is broken on i386
> :
> :    Support for atomic integer operations is/was broken on i386 platforms.
> :    The assembly code accidentally used opcodes that are only available on
> :    the i486 and later. So if you configured GCC to target, for example,
> :    i386-linux, but actually used the programs on an i686, then you would
> :    encounter no problems. Only when actually running the code on a i386
> :    will the problem appear.
> :
> :    This is fixed in 3.2.2."
>
> Well. Yes and no. It's wrong in the header file that should be
> used (the one for 386). But that won't help as the header file for 486
> is used.
>
> : But really it is not fixed. (may be in 3.2.3?)
>
> (Not yet.)

True. Not yet in GCC-3.3. At least even libstdc++ it is configured with

--target=i386-pc-msdosdjgpp --host=i386-pc-msdosdjgpp 
--build=i386-pc-msdosdjgpp

it uses i486+ instructions for atomic integer operations. I didn't 
reconfigure and rebuild entire gcc-3.3 though, only libstdc++-v3

Andris