Re: reporting/harness and stderr
Frank Lichtenheld <djpig@debian.org> writes:
> Yeah, that was the solution I thought off first but considered being
> too complex ;)
>
> Maybe a compromise might be the way to go:
> Since most of our stderr comes from programs we call, it might be
> easier to do the tagging for that in lintian itself and only handle
> stderr from the lintian process (e.g. Perl warnings) in harness.
Yeah, that could work. Most of the programs we run we already run through
helper routines. If we enhanced those using something like IPC::Run so
that we could more easily separately capture the stderr stream, we could
ignore the stuff we know we don't care about and issue tags for the rest.
It's a bit more complexity inside Lintian, but it may pay off in
robustness.
The hardest bugs to track down for me in the past (the current readelf
thing non-withstanding) have been Perl warnings from Lintian itself,
though. I'm not sure if setting $SIG{__WARN__} would let us do something
reasonable with those.
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: