[Debian-science-sagemath] Package cryptominisat, an optional dependency of SageMath

[Ximin Luo]

E. Madison Bray:
> On Tue, Nov 27, 2018 at 10:40 AM Ximin Luo <infinity0 at debian.org> wrote:
>>
>> Ximin Luo:
>>> Julian Rüth:
>>>> Hello Ximin,
>>>>
>>>> thanks a lot for having a look. This feedback was extremely helpful. I
>>>> uploaded a new version at
>>>> https://mentors.debian.net/package/cryptominisat that should address all
>>>> the things you pointed out. The only thing I didn't understand is the
>>>> FTBFS:
>>>>
>>>> * Ximin Luo <infinity0 at debian.org> [2018-08-11 07:20:00 +0000]:
>>>>> Your package FTBFS for me, there is some symbols mismatch:
>>>>>
>>>>> dpkg-gensymbols: warning: debian/libcryptominisat5-5.6/DEBIAN/symbols doesn't match completely debian/libcryptominisat5-5.6.symbols
>>>>> --- debian/libcryptominisat5-5.6.symbols (libcryptominisat5-5.6_5.6.4+dfsg.1-1_amd64)
>>>>> +++ dpkg-gensymbolsk7oqn0   2018-08-11 00:14:36.288737896 -0700
>>>>> @@ -8,13 +8,13 @@
>>>>>   _Z4calcPKSt6vectorIN5CMSat3LitESaIS1_EEbPNS0_16CMSatPrivateDataEb at Base 5.6.3
>>>>>   _Z5fromcPK5c_Lit at Base 5.6.3
>>>>>   _Z5fromcPK7c_lbool at Base 5.6.3
>>>>> - _ZN11MyOccSorterclERKN5CMSat7WatchedES3_ at Base 5.6.3
>>>>> - _ZN12StreamBufferIP8_IO_FILE2FNE8skipLineEv at Base 5.6.3
>>>>> +#MISSING: 5.6.4+dfsg.1-1# _ZN11MyOccSorterclERKN5CMSat7WatchedES3_ at Base 5.6.3
>>>>> +#MISSING: 5.6.4+dfsg.1-1# _ZN12StreamBufferIP8_IO_FILE2FNE8skipLineEv at Base 5.6.3
>>>>> [..]
>>>>> dh_makeshlibs: failing due to earlier errors
>>>>> debian/rules:7: recipe for target 'binary' failed
>>>>> make: *** [binary] Error 2
>>>>
>>>> I can not reproduce this. How exactly are you building this so I can try
>>>> to understand what might be the issue?
>>>>
>>>
>>> I couldn't reproduce this again either so I've run `dch -m -r` and uploaded your package to NEW. Thanks for your contribution to Debian!
>>>
>>
>> cryptominisat was recently accepted into Debian but these symbol difference errors are cropping up again on all buildds:
>>
>> https://buildd.debian.org/status/package.php?p=cryptominisat
>> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=914711
>>
>> Could you take a look please?
> 
> It's been a while since I've maintained a package for Debian, so I'm
> probably ignorant about something, but what is the problem if a few
> symbols changed?  Isn't that generally to be expected?  Especially in
> this sort of heavily templated C++ code, it's normal and expected that
> if some things were moved around, some template specializations may
> not be instantiated anymore, etc.  I'm not sure if that's what's going
> on here though.  Just curious.
> 
There may or may not be a "real problem" but often, symbols changes like this indicate a change in the ABI which causes problems when upgrading the library - other packages that runtime-depends on the library will be broken and require a re-build against the new library.

Since this package is a "new package" this really should not be a "real problem" but I am curious why the existing symbols file (as generated presumably by Julian) is incorrect today, but was OK a few months ago.

X

-- 
GPG: ed25519/56034877E1F87C35
GPG: rsa4096/1318EFAC5FBBDBCE
https://github.com/infinity0/pubkeys.git