Gentoo: dev-libs/gmp-4.2.1-r1 - TEST "t_import" fails with "-funroll-loops"/"-O3"

http://
bugs.gentoo.org/show_bug.cgi?id=180834


Mit freundlichen Grüßen
 
Steffen Bergner
 
- Abt. ODV -
--------------------------------

 
 
GGP Media GmbH
Karl-Marx-Straße 24
07381 Pößneck
Germany
www.ggp-media.de
 
steffen.bergnerbertelsmann.de
Fon:   +49 (0) 3647 - 430-285
Fax:   +49 (0) 3647 - 430-66 285
--------------------------------

GGP Media GmbH, Geschäftsführer: Christian Schantz
Sitz: Pößneck, Amtsgericht Gera-Stadt HRB 0481

Bitte beachten Sie auch die Rückseite dieses Schreibens.
_______________________________________________
gmp-bugs mailing list
gmp-bugsswox.com
http://swox
.com/mailman/listinfo/gmp-bugs

<Steffen.BergnerBertelsmann.de> writes:

  http://
bugs.gentoo.org/show_bug.cgi?id=180834
  
There isn't much the GMP developers can do about compiler
bugs.  We
used to try to work around them, but the compiler bug
bandwidth now
far exceed any sustainable workaround bandwidth, so we
stopped.

I actually suspect the set of correctly compiled non-trivial
C
programs is approaching the empty set, over the set of
compilers.  A
workaround might thereby just start triggering bugs in
another
compiler.

Our current approach to compiler bugs is to boom out to
users that
they should run "make check" and iterate over
several compilers until
they get lucky.  We trust the GMP test suite to catch any
relevant
compiler bug.

-- 
Torbjörn
_______________________________________________
gmp-bugs mailing list
gmp-bugsswox.com
http://swox
.com/mailman/listinfo/gmp-bugs

Hi Torbjorn,

thanks for your email.

I got it fixed for gcc 4.1.2. (-Wunsafe...), please have a
look into the bugzilla-case.

So "make check" runs fine after all. 

Thanks.

Regards,
 Steffen 

-----Original Message-----
From: tegeking.swox.se [mailto:tegeking.swox.se] On Behalf Of
Torbjorn Granlund
Sent: Monday, June 04, 2007 2:22 PM
To: Bergner, Steffen, GGP-KO
Cc: gmp-bugsswox.com
Subject: Re: Gentoo: dev-libs/gmp-4.2.1-r1 - TEST
"t_import" fails with
"-funroll-loops"/"-O3"

<Steffen.BergnerBertelsmann.de> writes:

  http://
bugs.gentoo.org/show_bug.cgi?id=180834
  
There isn't much the GMP developers can do about compiler
bugs.  We used to try to work around them, but the compiler
bug bandwidth now far exceed any sustainable workaround
bandwidth, so we stopped.

I actually suspect the set of correctly compiled non-trivial
C programs is approaching the empty set, over the set of
compilers.  A workaround might thereby just start triggering
bugs in another compiler.

Our current approach to compiler bugs is to boom out to
users that they should run "make check" and
iterate over several compilers until they get lucky.  We
trust the GMP test suite to catch any relevant compiler
bug.

--
Torbjörn
_______________________________________________
gmp-bugs mailing list
gmp-bugsswox.com
http://swox
.com/mailman/listinfo/gmp-bugs