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

Re: Introducing Build-Recommends / Build-Core-Depends?

Le Mon, Aug 15, 2011 at 11:12:36AM +0100, Steve McIntyre a écrit :
> Andreas Barth wrote:
> >
> >Generic options are usually better IMHO, but well - YMMV.
> Often, yes. But also often at extra cost. Where is the added benefit
> here - i.e. what are the use cases? I'm 100% behind making the
> bootstrap phase more simple, but I can't see many others...

I think Build-Recommends would be well suited for skipping or disabling
regression tests when ‘test dependancies’ are not available.  With
“DEB_BUILT_OPTIONS=nocheck”, the tests are skipped, but the packages needed to
run the tests are still installed, which means that if they are not available,
the package can not be built.  Here is an example:

bioperl-run contains BioPerl wrappers for common command-line tools.
t-coffee contains the T-COFFEE command-line tool to align nucleotidic sequences.

bioperl-run's regression tests try the wrapper for T-COFFEE, and therefore bioperl-run
Build-Depends on t-coffee. 

t-coffee fails to build from source on armel (where its regression test fails).

If t-coffee is removed from armel (where it probably never worked), bioperl-run
can not be rebuilt there.

This chain of build dependancies is weak, but brings the benefit of running
full regression tests at build time and include the results in the build logs.
I think that DEP 8 is complementary to this, but can not replace this feature
in the short term.

A Build-Recommends field would allow packages to be resilient to the absence of
one of the components featured in their regression tests. 

Have a nice day,

Charles Plessy
Debian Med packaging team,
Tsurumi, Kanagawa, Japan

Reply to: