Bug#825759: nmu: some Ada libraries, dw: to reflect order of build dependencies
On 21/06/16 23:41, Emilio Pozuelo Monfort wrote:
> On 31/05/16 17:10, Emilio Pozuelo Monfort wrote:
>> On 29/05/16 17:27, Nicolas Boulenguez wrote:
>>> Package: release.debian.org
>>> Severity: normal
>>> User: release.debian.org@packages.debian.org
>>> Usertags: binnmu
>>>
>>> Hello.
>>>
>>> Every package containing an Ada library embeds checksums of its
>>> recursive dependencies, in files suffixed with .ali. Everytime
>>> something attempts to build against the library, gcc checks that the
>>> sources of all dependencies match the checksum in the ALI files, else
>>> fails with a message about "obsolete ALI files".
>>>
>>> Between gcc-6.0 and gcc-6.1, the GNAT Ada runtime has changed. A
>>> library already built against gcc-6.0 and recursively depending on the
>>> modified part now contains an invalid checksum. All reverse
>>> dependencies are broken until the library is rebuilt against gcc-6.1.
>>>
>>> Ada libraries Build-Depend: gnat per policy, and gnat Depends: gnat-6
>>> (>= 6.1) since a few hours, so a bin-NMU will automatically wait for
>>> gcc-6.1 on architectures where it is not available yet.
>>>
>>> However, some dw commands are required to express that a library must
>>> only be rebuilt after its (library) dependencies.
>>>
>>> The already existing binNMU for libgnatcoll and asis on some
>>> architectures causes two version number to be involved, and the syntax
>>> in such an unusual case is not specified by the documentation. I hope
>>> that the intent is evident enough for any syntax issue to be easily
>>> fixed.
>>
>> There is no need to specify the +bN suffix in nmu commands (not sure if it's
>> just not necessary or actively harmful, but I've just removed them). For dw on
>> the -m switch it is correct.
>>
>>> For libaws on some architectures, no NMU is necessary because the
>>> build has always failed until now.
>>>
>>> Thanks for rescuing poor ALIs lost in Wonderland.
>>
>> All scheduled.
>
> Some of that failed because of the undefined references on libgnatprj6. That is
> fixed in 6.1.1-7, so I have given packages back (with appropriate dep-waits as
> gcc-6 is still building on some architectures yet) and things are looking well
> so far. Except for libaws on ppc64el (and sparc64), where it's complaining about
> some ali files. Are we missing some binNMUs? Can you take a look?
>
> https://buildd.debian.org/status/package.php?p=libaws
Also see asis/ppc64:
https://buildd.debian.org/status/package.php?p=asis
Emilio
Reply to: