Re: Bits from the autopkgtest maintainer
Antonio Terceiro wrote:
> On Mon, Mar 03, 2014 at 11:22:23PM +0100, Jakub Wilk wrote:
>> * Martin Pitt <mpitt@debian.org>, 2014-02-27, 17:38:
>> >Tests can now depend on "@builddeps@" which expands to the source
>> >package's build dependencies. This is useful if you have many
>> >build dependencies which are only necessary for running the test
>> >suite and you don't want to replicate them in the test Depends:.
>>
>> I advise to never use @builddeps@. It's a great way to shoot
>> yourself in the foot. :\
>
> would you ellaborate?
I would consider it to be a poor test of the built package if packages
that were not in Recommends, Suggests or packages specifically needed
as a test runner (e.g. python-nose) were being installed. Installing
packages other than those ones means that whether the
Depends/Recommends/Suggests are complete is not being tested [1]. This
is a class of bug that keeps coming up and one of the reasons why
autopkgtest tests are so attractive. There is of course another class of
tests where correct operation is ensured while other non-required packages
are installed (e.g. an alternative implementation etc) but that still isn't
build-deps.
(I don't know if that's what Jakub had in mind, but that's my 2¢)
cheers
Stuart
[1] Where possible, separate tests for Depends, Depends+Recommends,
Depends+Recommends+Suggests would seem even better to test the graceful
failure (or otherwise) in the absence of these related packages.
--
Stuart Prescott http://www.nanonanonano.net/ stuart@nanonanonano.net
Debian Developer http://www.debian.org/ stuart@debian.org
GPG fingerprint 90E2 D2C1 AD14 6A1B 7EBB 891D BBC1 7EBB 1396 F2F7
Reply to: