[devscripts] 01/01: move test clean up to test rule

[Nicholas Bamber]

Normally the debian stuff has to fill in whatever gaps are not handled 
well by the upstream makefiles. Often this includes cleaning up or 
moving files around. You have the luxury of being in control of upstream 
which means you can interpret "upstream" how you want. You choose to 
interpret upstream maximally but you could have chosen to interpret it 
minimally.

However I am NOT trying to persuade you to change. I just wanted to 
understand.

I have moved the clean up into the One time functions.

The emails I referred to were:
https://lists.alioth.debian.org/pipermail/devscripts-devel/2015-December/004370.html
https://lists.alioth.debian.org/pipermail/devscripts-devel/2015-November/004259.html

I am not really expecting an answer for either this side of the upload.

On 05/12/15 17:27, James McCoy wrote:
> On Sat, Dec 05, 2015 at 02:12:38PM +0000, Nicholas Bamber wrote:
>> James,
>> 	I can move it back of course.
>>
>> 	I am actually puzzled why more is not done in debian/rules.
>
> Like most packages, debian/ should only contain content that's relevant
> for packaging.
>
> Properly cleaning up detritus created by a test isn't specific to
> packaging, so it should be handled by the test (e.g., using shunit's
> oneTimeTearDown) or possibly by a clean target in test/Makefile.
>
>> The whole
>> devscripts structure seems sort of odd to me.
>
> Why?  Debian-specific stuff should be handled by things in debian/.
> "Upstream" stuff should be handled outside of debian/.
>
>> 	I put the test clean up in debian/rules because then one can easily inspect
>> the generated files between the test phase and the clean phase. More
>> generally I can think of several places such clean up could be put in:
>> 1. debian/rules
>> 2. Makefile
>> 3. test/Makefile
>> 4. test/test_*
>>
>> The two obvious places would seem to me 1. and 4. 4. seems not to be
>> working.
>
> Looking at test_package_lifecycle, it seems like use of either
> oneTimeSetUp/oneTimeTearDown or setUp/tearDown would help.
>
>> I am not quite sure why though I suspect the test script may be
>> exiting earlier than the cleanup. However 1. looks better to me anyway for
>> the reason I said. Both 2. and 3. look wrong to me. Why should test/Makefile
>> know about the inner workings of the tests?
>
> Why shouldn't it?  Do you expect other upstream Makefiles not to know
> how to clean up their generated files?  The use of “debian/rules clean”
> should typically only need to clean up files generated by the packaging,
> not by the upstream build process.  The upstream files should be cleaned
> by the upstream code.
>
>> BTW there are several quetsions I have asked which have not been answered.
>
> If they're buried in email chains about commits, where I wasn't already
> part of the thread, that's probably why.  Start new threads when you're
> discussing new topics.
>
>> Also what are the plans for upload?
>
> In the next week or so, probably.
>
> Cheers,
>