SWI-Prolog test failing (only!) on powerpc
Hi all,
I recently uploaded a new Debian package of ppl that had some minor packaging fixes; unfortunately, this time the build of the powerpc version fails with an error in the SWI-Prolog test suite:
--- ./../tests/expected_pgt 2009-04-18 05:03:19.000000000 +0000 +++ obtained_pgt 2010-02-07 14:08:37.000000000 +0000 @@ -1 +1,2 @@ +test for ppl_Octagonal_Shape_double_bounds_from_below failed. OK make[1]: *** [ppl_prolog_generated_test] Error 1 make[1]: Leaving directory `/home/mt/ppl-0.10.2/interfaces/Prolog/SWI' make: *** [check-am] Error 2
The full build log is available at [1]. I was able to reproduce this error on a powerpc system, but I have no idea how to debug this. The earlier builds used SWI-Prolog version 5.6.64, whereas this time 5.8.2 was used. I saw some fix related to 5.8 (9784b8165b6cb1dea5eb6d20b4cebca7ecc780fd) in git, but applying that did not have any effect on this problem. Furthermore, it seems to be fine on all other architectures.
Is there any other patch that I should have added?
Thanks, Michael
[1] https://buildd.debian.org/fetch.cgi?pkg=ppl&arch=powerpc&ver=0.10.2-...
Hi all,
I recently uploaded a new Debian package of ppl that had some minor packaging fixes; unfortunately, this time the build of the powerpc version fails with an error in the SWI-Prolog test suite:
[...]
For the moment I'll just work around this failure by mostly disabling this test. The attached patch includes this change plus an additional hack that was necessary to make PPL build on GNU/Hurd, which doesn't define SA_SIGINFO.
Best, Michael
On 02/13/10 22:12, Michael Tautschnig wrote:
The attached patch includes this change plus an additional hack that was necessary to make PPL build on GNU/Hurd, which doesn't define SA_SIGINFO.
Hi Michael,
thanks for the info. I have just pushed a change that should address the problem on GNU/Hurd:
http://www.cs.unipr.it/git/gitweb.cgi?p=ppl/ppl.git;a=commit;h=7ba6445b1c5ad...
Concerning the problem with SWI-Prolog, we would need a way to reproduce it. Cheers,
Roberto
On 02/13/10 22:12, Michael Tautschnig wrote:
The attached patch includes this change plus an additional hack that was necessary to make PPL build on GNU/Hurd, which doesn't define SA_SIGINFO.
Hi Michael,
thanks for the info. I have just pushed a change that should address the problem on GNU/Hurd:
http://www.cs.unipr.it/git/gitweb.cgi?p=ppl/ppl.git;a=commit;h=7ba6445b1c5ad...
Thanks a lot, I'll include that patch in the next upload instead of my own hack.
Concerning the problem with SWI-Prolog, we would need a way to reproduce it.
[...]
Does that mean that you don't have a PowerPC system at hand or just cannot reproduce it on any of those. Anyhow, this is a very strange problem as the very same PPL version worked fine in earlier builds. What has changed, however, is the SWI Prolog version. But as this is the same on other architectures as well, it must be some combination of PowerPC/SWI Prolog that causes this problem.
It's probably fine to just ignore this test and go ahead, and maybe changing the circumstances will make the problem go away by itself. Preferably, however, I'd debug it. My problem is that I don't really know where to start. I have finally understood how this test is generated, but is there some way of manually repeating the test in the Prolog interpreter? It's quite a while ago that I did my last steps in Prolog and back then it was just very little experiments. I tried to copy the test program line by line into the interpreter (to produce some more output at appropriate points), but then it already fails right after the first line (the "member" definition).
If I could get some directions how to debug the Prolog stuff (I'd love to understand how to that!) I can probably give you a lot more information about the actual root cause of this failure.
Best, Michael
participants (2)
-
Michael Tautschnig -
Roberto Bagnara