Re: [PPL-devel] New Parma Polyhedra Library packages
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.
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.
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. All the best, Roberto -- Prof. Roberto Bagnara Computer Science Group Department of Mathematics, University of Parma, Italy http://www.cs.unipr.it/~bagnara/ mailto:bagnara@cs.unipr.it
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.
participants (2)
-
Arthur Loiret -
Roberto Bagnara