Bug#544578: Lintian example hook causes pbuilder to purge *.deb if there are errors

Junichi Uekawa dancer at netfort.gr.jp
Tue Sep 8 14:34:44 UTC 2009


At Sun, 06 Sep 2009 01:04:17 +0800,
Chow Loong Jin wrote:
> 
> [1  <text/plain; UTF-8 (quoted-printable)>]
> On Sunday 06,September,2009 12:30 AM, Junichi Uekawa wrote:
> > At Thu, 03 Sep 2009 23:13:00 +0800,
> > Chow Loong Jin wrote:
> >>
> >> [1  <text/plain; UTF-8 (quoted-printable)>]
> >> On Thursday 03,September,2009 10:50 PM, Junichi Uekawa wrote:
> >>> Isn't it better to use C10shell instead of ignoring errors?
> >>>
> >> C10shell doesn't get invoked when a hook dies. Also, there are many cases where
> >> you don't want a shell to be spawned just because of a few lintian errors.
> >> Generally, at least for me, inspecting the resultant debs can already show
> >> what's wrong. But aside from that, anything that doesn't result in my built debs
> >> being purged due to lintian errors would satisfy me already. Perhaps pbuilder
> >> should instead be tweaked to not purge debs if a hook dies?
> > 
> > I find a behavior to not fail at all when lintian fails to be wrong,
> > that's ignoring lintian warnings for everybody; so your patch doesn't
> > look quite correct.
> If I'm not mistaken, B90lintian's purpose is to just print out the lintian
> warnings onto the build log output, right? The patch would enable the warnings
> to still be printed out, although it won't cause pbuilder to die a miserable death.

Here you are mistaken; the purpose of B90lintian is to fail the build
when there is a lintian error.

> > 
> > However, there should be a better behavior than the current behavior.
> > 
> > Should fixing C10shell to be invoked when hooks fail be useful?
> It might be, but I'd rather prefer B90lintian to not cause pbuilder to die a
> miserable death.
> 
> How about making pbuilder copy the resultant debs to $BUILDRESULT as long as the
> package builds fine, regardless of whether hooks die? Then perhaps using a
> certain exit code to show that a hook has died so that people can test the
> return code of pbuilder?

You don't want files in BUILDRESULT; you want to fix the bug; only the
good packages go through to BUILDRESULT.


> -- 
> Kind regards,
> Chow Loong Jin (GPG: 0x8F02A411)
> Ubuntu Contributing Developer
> 
> [2 OpenPGP digital signature <application/pgp-signature (7bit)>]
> 





More information about the Pbuilder-maint mailing list