Re: Test failures

[Russ Allbery <rra@debian.org>]

Raphael Geissert <atomo64+debian@gmail.com> writes:
> Frank Lichtenheld wrote:

>> Running test scripts-diversions 1.0... building... testing... FAILED:
>> --- t/tests/scripts-diversions/tags   2009-03-15 23:59:59.000000000 +0100
>> +++ debian/tests/tags.scripts-diversions      2009-03-28 16:42:09.000000000
>> +0100 @@ -2,4 +2,3 @@
>>  E: scripts-diversions: orphaned-diversion usr/share/scripts/orphan
>>  preinst E: scripts-diversions: package-uses-local-diversion preinst:12
>>  E: scripts-diversions: package-uses-local-diversion preinst:16
>> -E: scripts-diversions: remove-of-unknown-diversion
>> usr/share/scripts/old-file postrm:12 Running test binaries-from-other-arch
>> 1.0... building... testing... FAILED:

> No idea, but you are not alone:
> $ debian/rules runtests onlyrun=scripts-diversions
> [...]
> Running test scripts-diversions 1.0... building... testing... FAILED:
> --- t/tests/scripts-diversions/tags     2009-03-25 11:21:43.000000000 -0600
> +++ debian/tests/tags.scripts-diversions        2009-03-28
> 11:45:03.000000000 -0600
> @@ -2,4 +2,3 @@
>  E: scripts-diversions: orphaned-diversion usr/share/scripts/orphan preinst
>  E: scripts-diversions: package-uses-local-diversion preinst:12
>  E: scripts-diversions: package-uses-local-diversion preinst:16
> -E: scripts-diversions: remove-of-unknown-diversion
> usr/share/scripts/old-file postrm:12
> make: *** [runtests] Error 1

But it works fine here:

LINTIAN_ROOT="" /usr/bin/perl t/runtests -k t debian/tests scripts-diversions
Package tests:
Running test scripts-diversions 1.0... building... testing... ok.

I'm mystified, since I can't see anything about this check that would
change based on what system it was run on, or any remaining ordering
problems with what order in which the maintainer scripts were read.

I did a fresh git clone on a different system and it worked fine there as
well.

>> --- t/tests/binaries-from-other-arch/tags     2009-03-15 23:59:59.000000000
>> +0100
>> +++ debian/tests/tags.binaries-from-other-arch        2009-03-28
>> 16:42:57.000000000 +0100 @@ -1,4 +1,4 @@
>>  E: binaries-from-other-arch: binary-from-other-architecture
>>  ./usr/bin/elfobject E: binaries-from-other-arch: statically-linked-binary
>>  ./usr/bin/elfobject
>> -W: binaries-from-other-arch: apparently-truncated-elf-binary
>> ./usr/bin/elfobject +W: binaries-from-other-arch:
>> apparently-corrupted-elf-binary ./usr/bin/elfobject
>>  W: binaries-from-other-arch: binary-without-manpage usr/bin/elfobject

> This one is tricky; you need squeeze's binutils for it to pass.

Works fine for me on both i386 and amd64.  Frank, do you have an older
binutils?

>> Running test files-windows-devel-file-in-package 1.0... building...
>> testing... FAILED:
>> --- t/tests/files-windows-devel-file-in-package/tags  2008-11-28
>> 12:06:02.000000000 +0100
>> +++ debian/tests/tags.files-windows-devel-file-in-package     2009-03-28
>> 16:48:28.000000000 +0100 @@ -1 +1,2 @@
>> +I: files-windows-devel-file-in-package source:
>> debian-watch-file-is-missing
>>  W: files-windows-devel-file-in-package: windows-devel-file-in-package
>>  usr/lib/foo/foo.vcproj
>> Running test scripts-ocamlrun 1.0... building... testing... FAILED:
>> --- t/tests/scripts-ocamlrun/tags     2008-11-28 12:06:02.000000000 +0100
>> +++ debian/tests/tags.scripts-ocamlrun        2009-03-28 16:50:22.000000000 +0100
>> @@ -1 +1,2 @@
>> +I: scripts-ocamlrun source: debian-watch-file-is-missing
>>  W: scripts-ocamlrun: binary-without-manpage usr/bin/script1

> The watch files should be being created by runtests, so it is a bit
> odd.

They won't be since those are native packages.  The question is why
Lintian is treating them like non-native packages and expecting a watch
file, but not producing any of the other tags that would fit it thinking
they're non-native packages (like bad version number).

if (! -f "debfiles/watch") {
    tag 'debian-watch-file-is-missing' unless ($info->native);
    return;
}

is the test, so unless $info->native is broken somehow for source packages
on Frank's system but not on ours, I'm not sure what's going on.

Currently, for version 1.0 packages, $info->native on a source package
checks for the existence of the *.diff.gz file in the lab.  Is something
causing this to break for some reason?

        my $version = $self->field('version');
        $version =~ s/^\d+://;
        my $name = $self->{name};
        $self->{native} = (-f "${name}_${version}.diff.gz" ? 0 : 1);

-- 
Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>