[Debian-med-packaging] Bug#860655: Exclicitly or "implicitly" mark architectures a packages does not build

[Andreas Tille]

Hi,

in spring several packages from Debian Med team received "FTBFS on i386:
unsatisfiable build-dependencies" bug reports (sorry for not caring
earlier).  An "example bug" #860655 is in CC.  Originally these bugs
were filed with severity serious but at it was pointed out by Gianfranco
Costamagna[1] the severity of this issue for packages that were never
built on a certain architecture are not serious.  So the severity of all
these bugs was decreased to wishlist.

Since we also intend to care for wishlist bugs I'm now wondering the
following:  Yes, there are packages that do not build on a certain
architecture due to missing Build-Depends.  I could exclude these
architectures from the list of architectures in d/control.  However, as
far as I understood that's rather a bad idea since once the
Build-Depends might become available the package could easily build.

In some cases we even added a Build-Depends which actually is not really
a Build-Depends but just a Depends of the package but there is no point
to build the package if it can not be installed afterwards (see for
instance gasic[2]).

So as far as I see #860655 there is no sensible means to fix it properly
and I'm tempted to close it (since I also do not see any sense in
tagging it wontfix).  May be it needs to be said that we do not have the
manpower to fix each and every piece of code to make sure all
Build-Depends build on every architecture neither does it make sense
technically to for instance have gene sequencing software on outdated
hardware available.  Some code just needs 64 bit and upstream will not
support other hardware.

Am I missing something?

Kind regards

       Andreas.

[1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=860652#15
[2] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=873859#15

-- 
http://fam-tille.de