Re: can buildd logs be sorted (again)?
Thomas Viehmann wrote:
> Hi,
>
> Steve M. Robbins wrote:
>> The buildd log pages, e.g. [1], used to be sorted by package version
>> (or maybe build date). However that is no longer the case.
>
> It seems that build.php relies on the results of opendir/readdir to be
> in order and that assumption not being valid anymore.
>
>> Can this be fixed? The current situation is less than useful since
>> the latest build is buried in other output.
>
> If someone could attest to
> http://buildd.debian.org/~tviehmann/cmp_versions_php.txt
> (translated from the perl that udd uses) being wrong in only harmless
> ways (except showing that C->Perl->PHP with the last step being done by
> someone disliking PHP is not a good idea), I could ask Ryan to put
> http://buildd.debian.org/~tviehmann/build.php
> in the appropriate location.
Nice, yet another rewrite of dpkg's code, but now in PHP[1].
So, let's test it with a script I wrote during DC8 (as suggested by Lucas):
8<------------------------------------------------------------------------->8
buildd.php failed test 'foo- foo 0'/tests/foo with '1'
buildd.php failed test 'foo fo 1'/tests/foo with '-1'
buildd.php failed test 'foo- foo+ -1'/tests/foo with '1'
buildd.php failed test 'foo~1 foo -1'/tests/foo with '1'
buildd.php failed test 'foo.bar foo1bar 1'/tests/foo with '-1'
buildd.php failed test 'foo.bar foo0bar 1'/tests/foo with '-1'
buildd.php failed test '1foo:bar 1:bar x This is a syntax violation so the
result is "undetermined" but shouldn't be any of these: -1 0 1'/tests/foo
with '-1'
buildd.php failed test '1foo-1 foo-1 -1'/tests/foo with '1'
buildd.php failed test '0.8.0a-1 0.8.0-2 1 gnunet'/tests/versions.lenny-sid
with '-1'
buildd.php failed test '0.8.0a-1 0.8.0-1 1 gnunet-gtk'/tests/versions.lenny-sid
with '-1'
buildd.php reached the 10 tolerated failures
The remaining tests will be skipped
8<------------------------------------------------------------------------->8
With regard to the '1foo:bar 1:bar' test: this test is usually passed when the
versions comparing function returns null (in PHP terms), making the 'case'
script exit with a non-zero status. In your implementation I see parseversion
returning null when it encounters an invalid version, but cmp_version doesn't
care about it.
The tests are in a "version1 version2 expected_result comments"-format.
I will later test udd's perl implementation just to make sure it is ok (although
I believe it's PET's which is not perfect either).
[1] I promised Christoph Berg that I am going to write a PHP extension, actually
an interface, to libapt's functions; but I haven't been lucky enough to find
the docs (libapt-pkg-doc isn't what I expected). If somebody can point me to
the right documents it would be great.
>
> Kind regards
>
> T.
Cheers,
Raphael Geissert
Reply to: