[Debian-med-packaging] backporting libssw - autopkgtest fails on jessie

[Sascha Steinbiss]

Hi Afif,

> Ok. I haven't uploaded yet (and I'm all done with my Debian time for
> now) because I see these lintian issues that I don't think you have in
> your version from unstable:
> 
> I: libssw changes: backports-changes-missing
> W: libssw source:unknown-architecture any-x32
> I: libssw0: no-symbols-control-file usr/lib/x86_64-linux-gnu/libssw.so.0
> W: libssw0: maintscript-calls-ldconfig postinst
> E: libssw0: package-must-activate-ldconfig-trigger
> usr/lib/x86_64-linux-gnu/libssw.so.0
> W: libssw0: maintscript-calls-ldconfig postrm
> I: libssw-dev: static-library-has-unneeded-section
> usr/lib/x86_64-linux-gnu/libssw.a(ssw.o) .comment
> I: libssw-dev: static-library-has-unneeded-section
> usr/lib/x86_64-linux-gnu/libssw.a(ssw_cpp.o) .comment
> I: ssw-align: hardening-no-pie usr/bin/ssw-align

No, I for sure I don’t have any errors. I have fixed the unknown-architecture warning in the meantime though. The rest seems to be concerned with ldconfig calls - IIRC this is not needed to be done manually?

> It's possibly due to the backports lintian version being slightly
> outdated (2.5.44~bpo8+1 vs 2.5.45) but I haven't checked yet. Just
> letting you know why I haven't uploaded.

OK, no problem.

>>> I'd prefer this way to avoid in Jessie the performance penalty you
>>> mentioned that comes with 1.0-4.
>> 
>> Well, it's not clear yet whether this really makes a difference, but it
>> could -- from my test runs I haven't seen any indication for that but
>> I'd rather let upstream decide.
>> 
>> BTW, SPAdes fails during linking when I give it my new 1.0-4 in a local
>> apt repo. I'll need to investigate whether this is connected to my
>> changes or not, but it'll have to wait until later today.
> 
> Hmm. This package has had some interesting issues.

Yes, indeed. Taking another look at my changes to fix the results corruption issue, I noticed that the change subtly changed the API — a function that used to be implemented in the header files for inlining is now exported from the library itself. This also caused the problems building SPAdes against that version (as it still used the old headers from the included copy of libssw — I just fixed that).
I must admit that I’d rather wait with uploading my fixed version until upstream have agreed on a solution, I don’t want to unnecessarily cause an ABI bump when I have to remove the function from the library again if upstream decide on a different fix.

Cheers
Sascha
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 455 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.alioth.debian.org/pipermail/debian-med-packaging/attachments/20160719/1c6109fa/attachment-0001.sig>