[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: <math.h> on m68k and pari



On Thu, Jun 29, 2006 at 10:16:47AM -0500, Stephen R Marenka wrote:
> On Thu, Jun 29, 2006 at 04:00:45PM +0200, Bill Allombert wrote:
> 
> > As you know, due to bug #356045 in libc6-dev, it is not possible to
> > compile a program that #include <math.h> with -O and above.
> > 
> > This behaviour occurs in particular in the sid chroot of crest, making
> > impossible to test-build some packages there.
> > 
> > I would like to investigate the build failure of pari (2.3.0-1) on m68k.
> > (I have to say the packages was building fine with gcc 4.1.0-2 so
> > I am surprised that gcc 4.1.1-2 miscompile it.
> 
> Are you sure? This bug has been in glibc for a while, certainly before I 

Yes, but #356045 has been closed and then reopened instead of opening a
new bug. The fix of the initial #356045 has a typo that cause the 
current bug.

echo "#include <math.h>" > foo.c
gcc -O foo.c
In file included from /usr/include/math.h:382,
                 from foo.c:1:
/usr/include/bits/mathinline.h:161: error: expected ',' or ';' before '{' token
/usr/include/bits/mathinline.h:161: error: expected ',' or ';' before '{' token

% dpkg -s libc6-dev
Package: libc6-dev
Version: 2.3.6-15

> > In the meantime I have uploaded pari (2.3.0-2) which use a lower 
> > optimisation on m68k, but I still want to investigate the issue.
>  
> In any case, Roger Leigh has a working patch which should show up in the
> next glibc. It's listed at the end of the bug report. Please feel free
> to get with him and make sure this works for everyone.

I know, but I think we should try to get the fixed glibc uploaded asap
because a breakage in math.h is a rather serious problem.

Cheers,
-- 
Bill. <ballombe@debian.org>

Imagine a large red swirl here. 



Reply to: