[lug] povray on fedora core 4

Zan Lynx zlynx at acm.org
Mon Sep 12 21:44:27 MDT 2005


I just tested a build on x86_64, with GCC 4.0.1 and had no errors.  Why
don't you post the machine instruction that it failed on?  You should be
able to use GDB and after it dies, enter "disass" and look for the
instruction at the address where it failed.

I suspect that you did something like set the wrong march or mcpu
options, or you're using a weird CPU that looks like i686 but isn't
really, like some of the Cyrix chips.  Did you tell GCC to
auto-vectorize the code?  Does the CPU support SSE2?

Actually, just post what CFLAGS were used to build, what the CPU is, and
the disassembly of the instruction where it dies.

On Mon, 2005-09-12 at 19:29 -0600, D. Stimits wrote:
> I'm trying to build povray on fedora core 4 (x86), and running into a 
> problem I've never seen before. I think this is probably a gcc4 bug. I'm 
> curious if anyone here with fedora core 4 or gcc 4 has successfully 
> built povray 3.6.1? See:
> http://www.povray.org/download/
> ftp://ftp.povray.org/pub/povray/Official/Linux/
> 
> The very very unusual thing that I see is that it builds fine, but any 
> attempt to run it results in a core dump...not just any core dump, but:
> Illegal instruction (core dumped)
> 
> When I rebuild this with debug symbols, I find the core dump for illegal 
> instruction occurs in rendctrl.cpp, on this line:
> opts.First_Column_Percent = 0.0;
> 
> This is extremely weird, because there is no way this code should be a 
> problem, unless gcc itself is generating invalid code. The system itself 
> is very stable, and I have found (and reported) other bugs which were 
> verified. Now my question is this...can anyone else with gcc4 or fedora 
> core 4 create this same error? You'd have to compile povray 3.6.1, then 
> run it with any argument (or none at all), since the core dump is during 
> init stuff. I want to find out if this happens to anyone else. The 
> povray people have said they believe it is a compiler problem, and I 
> tend to agree...but if I can't verify it then I can't turn in a bug 
> report to gcc.
> 
> D. Stimits, stimits AT comcast DOT net
> _______________________________________________
> Web Page:  http://lug.boulder.co.us
> Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug
> Join us on IRC: lug.boulder.co.us port=6667 channel=#colug
-- 
Zan Lynx <zlynx at acm.org>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://lists.lug.boulder.co.us/pipermail/lug/attachments/20050912/16f2f52b/attachment.pgp>


More information about the LUG mailing list