Bug#847168: devscripts: debuild fails when lintian fails, regression since 2.16.9

[James McCoy]

On Tue, Dec 06, 2016 at 09:16:33AM +0100, Raphaël Hertzog wrote:
> Since debuild now relies on dpkg-buildpackage's hook to run lintian,
> a failing lintian fails the whole build process. This was not the case
> before 2.16.9. So this is either a regression or a annoying new feature.

It was an unforeseen change, but one that I think is an improvement (as
noted in #846192).

> When I work on Kali packages, I almost always get a lintian error because
> lintian doesn't know of the suite

Can't that particular issue be fixed with vendor information provided
for lintian?  That's already done for Ubuntu.

> or because my name does not appear in
> the maintainer field.

This can also be disabled in general for Kali, if that's the expected
behavior.

> I want to be informed of the errors but I don't
> want the whole build process to be stopped... I want "gbp buildpackage" to
> create my tag and I want debuild to sign my package.

While I understand that the change is unexpected, wouldn't it be better
to resolve the errors "globally" rather than just for your dev
environment?

> If you consider this a new feature, then I ask you to document how to
> configure debuild to not fail on lintian failures (possibly adding a new
> option to ignore the result of the check-command).
> 
> If you consider this a regression, then the default check hook should have
> some "|| true" added at the end or something like that (or you can also
> add a new option to ignore the result and have its default to true instead
> of false).

I can't do any of the above without pulling the execution of lintian
back into debuild.  The actual command-line being run isn't in debuild's
control (to the extent that we could do an equivalent of "|| true") and
all debuild knows is that dpkg-buildpackage failed.  It has no idea that
it was due to the check command (lintian).

Cheers,
-- 
James
GPG Key: 4096R/91BF BF4D 6956 BD5D F7B7  2D23 DFE6 91AE 331B A3DB