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

Re: New Parma Polyhedra Library packages



Hi Roberto,

On Tue, Sep 23, 2008 at 09:42:16PM +0200, Roberto Bagnara wrote:
> Matthias Klose wrote:
> >Hi, we did need the packages as a build dependency for gcc-snapshot
> >(trunk).  the testsuite is run, but we don't abort the build on
> >errors. see the build logs at http://buildd.debian.org/build.php?pkg=ppl,
> 
> Hi Matthias,
> 
> I have checked the build logs, but `make check' does not seem
> to be run.  In contrast, `make world' is run into the `doc'
> subdirectory: this generates tons of developer references
> that I doubt are of any interest to the ordinary user.
> If you can let us know which kind of documentation you
> want to include in the package, we can advise.
> 
> Concerning tests, I think it is very important that `make check'
> is run on each architecture to make sure there are no problems:
> perhaps not on every build (since it is a very heavy procedure),
> but at least before any release.

The testsuite will be run on next upload (0.10~pre27-4) on all architectures.
And the generated documentation is in a separated libppl-doc package, so it
should be ok.


> >current bug reports at http://bugs.debian.org/src:ppl.
> 
> I had a look at them: I have fixed one in CVS (the next PPL 0.10
> snapshot will come in a few days). For the other bug, I have offered help.
> 
> In general, at least for the PPL, reporting the bugs upstream is
> the quickest way to have them fixed.

Thanks for this! I will try to report Debian bugs upstream from now.


> >I hope that Arthur Loiret can provide access to an alpha machine. Note
> >that Debian conigures gcc on alpha to include -mieee by default. for
> >testing you'll find other gcc versions installed on this machine
> >(gcc-4.x).
> 
> Notice that -mieee is not enough: according to the GCC documentation
> you need -mieee-with-inexact, and even that is not enough
> to obtain full IEEE 754 compliance (see
> http://gcc.gnu.org/ml/gcc/2008-09/msg00400.html).
> We are testing right now a new configuration procedure that
> should be able to detect and work around such bogus implementations of
> floats.
> 
> Concerning the access to machines, thanks to FSF Europe we have
> access to an Alpha.  What we miss, is access to the following:
> ia64, hppa, arm, s390, mips, mipsel, sparc (but we have sparc64),
> armel, and m68k.

The sparc hardware (gcc30) is running a sparc64 kernel with a 32-bit
userspace, so you can either run gcc (or gcc -m32) to generate 32-bit code,
or gcc -m64 to generate 64-bit code. Both can run on the system, but keep in
mind 32-bit is the default.


Arthur.

Attachment: signature.asc
Description: Digital signature


Reply to: