[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#807849: nghttp2: cross Build-Depends unsatisfiable



On 15/12/15 07:11, Helmut Grohne wrote:
> Control: tags -1 - patch
> 
> [...]
> 
> I'm sorry for having screwed this up. Indeed the multiarch spec and the
> version of dpkg in archive do not allow annotating arch:all packages
> with :native. This issue is still being discussed with dpkg and
> multiarch maintainers and doesn't look like it is being solved quickly.
> 
> There are two ways to fix this (applying to both jansson and nghttp2):

Hi Helmut,

> 
> 1) Since the documentation resides in Arch:all -doc packages, arguably
>    python-sphinx is not needed for building architecture dependent
>    packages. Thus the dependency should be moved to Build-Depends-Indep.
>    Of course I tried this method on both packages before resorting to
>    the :native annotation, but the use of sphinxdoc in debhelper's
>    --with flag makes this non-trivial. If you see a way to do this,
>    that'd totally solve the issue.
> 
> 2) Turn python-sphinx into an architecture dependent package (i.e.
>    convert from Arch:all to Arch:any). The :native annotation is allowed
>    on architecture dependent packages and thus (after python-sphinx is
>    converted) the patches I sent can be used as is. (This is essentially
>    the "multiarch interpreter workaround" in the "Automating
>    Architecture Bootstrap" DC15 talk.)
> 
> Tomasz, do you have a preference for nghttp2? I'll try to work that out.
> 
> I have a slight preference for the former, because I generally prefer
> having Arch:all packages over lots of identical Arch:any packages. At
> the same time I acknowledge that it seems likely that we have to do the
> conversion anyway as not all reverse build-depends of python-sphinx
> separate their documentation to architecture independent packages as
> jansson and nghttp2 do.
> 
> Helmut

Ok, I'll try to implement 1) and let you know about my success/fail.

Tomasz

Attachment: signature.asc
Description: PGP signature


Reply to: