[Debian-science-sagemath] Fwd: Bug#888606: cysignals: FTBFS with debhelper/11.1 due to empty build target

[Tobias Hansen]

I still get a segfault when testing src/sage/structure/sage_object.pyx. Do you also get that with 8.1-3? If not we can upload.

Best,
Tobias

On 02/07/2018 04:19 PM, Ximin Luo wrote:
> I tested 8.1-2 against latest unstable and these errors have gone away.
>
> I see you made some changes for 8.1-3 which fixes some other issues (matplotlib etc), they look good to me, I'll let you upload that when you're ready.
>
> X
>
> Tobias Hansen:
>> Yes, looks like it might be fixed in the last python upload:
>>
>>  python2.7 (2.7.14-6) unstable; urgency=medium
>>  .
>>    * Don't run lib2to3 tests which rely on the pickled grammar files.
>>
>>
>>
>> On 02/06/2018 09:37 PM, Ximin Luo wrote:
>>> giac is now fine, sagemath still failing.
>>>
>>> https://buildd.debian.org/status/fetch.php?pkg=sagemath&arch=ppc64el&ver=8.1-2%2Bb2&stamp=1517948950&raw=0
>>>
>>>         python_grammar = driver.load_packaged_grammar("lib2to3", _GRAMMAR_FILE)
>>>     AttributeError: 'module' object has no attribute 'load_packaged_grammar'
>>>
>>> Probably due to recent python changes since the last upload. I'll look tomorrow/day after, and also package giac 1.4 (or feel free to jump in ahead too).
>>>
>>> X
>>>
>>> Ximin Luo:
>>>> At least flint-arb looks like MPFR4 issues, another version was released recently. Sucks that all these "bugs" were filed against our packages before more detailed investigation about the cause.
>>>>
>>>> I've been away from home, will take a closer look when I get back in the next few days.
>>>>
>>>> X
>>>>
>>>> Tobias Hansen:
>>>>> Hi all,
>>>>>
>>>>> I could not reproduce the FTBFS of giac on amd64. Should I just reupload giac with the unimportant change that is in git to see if it works now? Or upload the new version 1.4.9.45? (sagemath 8.2.beta4 uses this version.)
>>>>>
>>>>> However with giac built against libmpfr6 and libmpfr4 uninstalled I still get the same segfaults with sagemath.
>>>>>
>>>>> flint-arb and cypari2 also have problems building. :(
>>>>>
>>>>> Best,
>>>>> Tobias
>>>>>
>>>>>
>>>>>
>>>>> On 01/29/2018 02:22 PM, Tobias Hansen wrote:
>>>>>> Ah yes, giac is still using libmpfr4 because its rebuild failed on amd64.
>>>>>>
>>>>>> Best,
>>>>>> Tobias
>>>>>>
>>>>>> On 01/28/2018 01:38 PM, Ximin Luo wrote:
>>>>>>> I saw some issue, I forgot where, that pointed out both mpfr4 and mpfr6 were getting pulled into certain builds. You could check your sbuild logs to see if both are getting installed.
>>>>>>>
>>>>>>> X
>>>>>>>
>>>>>>> Tobias Hansen:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> yes, I think you can update cysignals in unstable. I applied upstream patches for new mpfr and matplotlib versions to sagemath and built it with both versions of cysignals yesterday and got the same problems:
>>>>>>>>
>>>>>>>> sage -t --long src/sage/rings/number_field/number_field.py  # Bad exit: 1
>>>>>>>> sage -t --long src/sage/rings/polynomial/plural.pyx  # Killed due to segmentation fault
>>>>>>>> sage -t --long src/sage/structure/factorization_integer.py  # Killed due to kill signal
>>>>>>>> sage -t --long src/sage/structure/sage_object.pyx  # Killed due to abort
>>>>>>>>
>>>>>>>> I have had the failure with number_field.py on my computer for a while, but it doesn't happen on buildds. The others are new, but I don't even know if the libmpfr6 transition is finished...
>>>>>>>>
>>>>>>>> Best,
>>>>>>>> Tobias
>>>>>>>>
>>>>>>>>
>>>>>>>> On 01/28/2018 06:11 AM, Jerome BENOIT wrote:
>>>>>>>>> Hello,
>>>>>>>>>
>>>>>>>>> this issue was fixed in 1.6.6+ds-2 which is current in experimental:
>>>>>>>>> can we migrate cysingals 1.6.6+ds-2 to unstable ?
>>>>>>>>>
>>>>>>>>> Thanks,
>>>>>>>>> Jerome
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> -------- Forwarded Message --------
>>>>>>>>> Subject: Bug#888606: cysignals: FTBFS with debhelper/11.1 due to empty build target
>>>>>>>>> Resent-Date: Sat, 27 Jan 2018 17:51:02 +0000
>>>>>>>>> Resent-From: Niels Thykier <niels at thykier.net>
>>>>>>>>> Resent-To: debian-bugs-dist at lists.debian.org
>>>>>>>>> Resent-CC: Jerome Benoit <calculus at rezozer.net>
>>>>>>>>> Date: Sat, 27 Jan 2018 18:44:34 +0100
>>>>>>>>> From: Niels Thykier <niels at thykier.net>
>>>>>>>>> Reply-To: Niels Thykier <niels at thykier.net>, 888606 at bugs.debian.org
>>>>>>>>> To: Debian Bug Tracking System <submit at bugs.debian.org>
>>>>>>>>>
>>>>>>>>> Source: cysignals
>>>>>>>>> Version: 1.6.5+ds-2
>>>>>>>>> Severity: serious
>>>>>>>>> Tags: patch
>>>>>>>>>
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> The cysignals package FTBFS with debhelper/11.1 as it has an empty
>>>>>>>>> build target.  This is caused by debhelper had a bug in its handling
>>>>>>>>> of "explicitly defined rules targets" that has now been fixed.
>>>>>>>>>
>>>>>>>>> Previously, this happened to work because dpkg-buildpackage would
>>>>>>>>> invoke "debian/rules build" (which would be a no-op) followed by
>>>>>>>>> "fakeroot debian/rules binary".  During the binary target, dh's
>>>>>>>>> suboptimal handling would run the build commands.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> The solution is trivial but less pretty; explicitly define "build"
>>>>>>>>> with the same content as the "%:" target (or rename the "build" folder
>>>>>>>>> and drop the ".PHONY" target).  I have attached a patch for this.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> More details can be found in:
>>>>>>>>>  * #886901 comment #35
>>>>>>>>>  * #887688 comment #37
>>>>>>>>>  * #880840
>>>>>>>>>
>>>>>>>>> Apologies for the inconvenience.
>>>>>>>>>
>>>>>>>>> Thanks,
>>>>>>>>> ~Niels
>>>>>>>>>
>>>>>>>>>