On 2011-06-28 19:02, Russ Allbery wrote: > Niels Thykier <ni...@thykier.net> writes: > >> For this I was wondering if > >> my $job = { err => 'index-errors', ... } > >> appends the output to 'index-errors' or truncates the file before >> writing. > > It truncates currently. Lintian::Command could be modified to make it > append, though, which may not be a bad idea. I think all one would have > to do is change 2> to 2>>. >
That was easy to do - I decided to implement this by using a new option for it instead of changing the behaviour of out/err. Makes life less interesting. :) >> Alternatively, would we rather want cruft to handle multiple index-error >> files (one for each tarball)? > > I'm not sure it's worth the effort. Errors there happen rarely. > I was not sure either, but I wanted to mention a possible alternative. >> Secondly, if I understand the situation correctly, I believe the version >> of collection/index.desc should be updated with this change (to ensure >> the index are re-generated). But since file-info generates its output >> from the index file, file-info should be re-run as well. >> Unfortunately, if I understand frontend/lintian correctly this will >> not happen automatically. Since file-info succeeded in the past, there >> should already be a ".file-info-1" file in the lab and the update of >> index will not make frontend/lintian remove that (or have I missed >> something?). > > I think that's correct. Collection scripts should probably check the > version placeholders of all their dependencies as well, but I don't think > they currently do that. > Personally I sort of figured our make-shift resolver ought to deal with that. Especially since the collections are being run in parallel. ~Niels -- To UNSUBSCRIBE, email to debian-lint-maint-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/4e0afdf2.9000...@thykier.net