[Pkg-privacy-maintainers] MAT 0.6 packaging

[Sascha Steinbiss]

Hi intrigeri,

> I had a closer look, and uploaded the package! :)

Thanks for the review! :)

> I hadn't noticed that you had *moved* the test suite run to
> autopkgtest. I can see that it's an optimization when preparing a new
> package (one doesn't have to run the test suite twice), but I'm not
> sure I really like this change. IMO, build-time tests and autopkgtests
> serve slightly different purposes, e.g. one can't assume all uploaders
> will run the autopkgtests prior to upload, and with this change, the
> full archive rebuilds done by QA people would not run the test suite
> anymore. So, I'd like to keep running the test suite via
> override_dh_auto_test too, done in commit d400b30 for the 0.6.1
> upload, but of course you may disagree and we can discuss it if
> needed :)

That's absolutely fine! My reasoning there was that what ultimately
matters is how the package performs in its installed state, and that any
test checking end user functionality run at build time is probably not
sufficient.
But it sure can't hurt to run the tests at build time as well so the
build fails if the test fails. So no objection ;)

> The autopkgtest fail for me (I'm using the QEMU backend, didn't try
> LXC):
> 
>   test-installed       FAIL stderr: test_remove (clitest.TestRemovecli)
> 
> Maybe it's fine and we should tell the autopkgtest that the test suite
> is expected to send stuff to STDERR?

Interesting... I didn't get that. I'd be surprised if the success of the
autopkgtest depends on whether a _testbed_ emits stuff on stderr or not.
I'll try it on qemu too check if a 'allow-stderr' helps.

> As some other team-members can confirm, I have a serious problem with
> spurious trailing whitespace at EOL, and I can't help insisting that
> it should not exist. I've removed some newly introduced ones with
> commit 7478985. It would make my life simpler if you configured your
> Git environment to notice such things more easily :)

Hehe, actually I agree! That's why I indeed colour them in the diffs:

$ cat ~/.gitconfig | grep whitespace
  whitespace = red reverse

This one must have slipped through. I will add a local pre-commit hook
checking for such cases to be sure.

Cheers
Sascha


-- 
 The Wellcome Trust Sanger Institute is operated by Genome Research 
 Limited, a charity registered in England with number 1021457 and a 
 company registered in England with number 2742969, whose registered 
 office is 215 Euston Road, London, NW1 2BE.