Bug#846711: git-build-recipe: FTBFS: Tests failures

[Guillem Jover]

Hi!

On Wed, 2016-12-07 at 09:14:36 +0000, Colin Watson wrote:
> On Tue, Dec 06, 2016 at 10:54:21PM -0500, James McCoy wrote:
> > On Tue, Dec 06, 2016 at 10:39:44AM +0000, Colin Watson wrote:
> > > git-build-recipe doesn't do anything particular to ask debuild to run
> > > lintian here, and it doesn't expect or require a lintian check.  It used
> > > to be that debuild would check whether lintian was in fact installed,
> > > and not run it if it wasn't; indeed, its documentation still says "then
> > > runs lintian on the .changes file created (assuming that lintian is
> > > installed)", thereby claiming that that's still what it does.  But this
> > > was broken in devscripts 58eb4a4a5e006bf9a2589da0ef2f36aa0d81ed8c when
> > > changing debuild to use dpkg-buildpackage --check-command.
> > 
> > Indeed.  In order to fix this, I either need to move running lintian
> > back into debuild or dpkg needs to not error when the check command
> > isn't present.
> 
> I don't understand why.  Wouldn't something like this (untested) be
> sufficient?
> 
> diff --git a/scripts/debuild.pl b/scripts/debuild.pl
> index 639bbac..afbd1dc 100755
> --- a/scripts/debuild.pl
> +++ b/scripts/debuild.pl
> @@ -988,7 +988,7 @@ else {
>      open STDOUT, ">&BUILD" or fatal "can't reopen stdout: $!";
>      open STDERR, ">&BUILD" or fatal "can't reopen stderr: $!";
>  
> -    if ($run_lintian) {
> +    if ($run_lintian and system("command -v lintian >/dev/null 2>&1") == 0) {
>  	push(@dpkg_opts, '--check-command=lintian',
>  	    map { "--check-option=$_" } @lintian_opts);
>      }

Yes, please. I'da rather not change the current behavior in
dpkg-buildpackage, because I think that if the user explicitly asked
to run a command that that does not exist, the correct thing to do is
to fail stating so. :)

Thanks,
Guillem