Re: [Module-build-general] RFC: Build test test_dirs=(list)

[Steve P. <sp...@qu...>]

On Wednesday, August 13, 2003, at 01:02  am, Michael G Schwern wrote:

> On Tue, Aug 12, 2003 at 11:51:16PM +0100, Steve Purkis wrote:
>>> If I want to test only the methods in the parent class Foo::Bar::One,
>>> I run "Build test test_dirs=t/Foo/Bar/One/t".   If I want to run
>>> the tests for that class and all it's children, I run
>>> "Build test test_dirs=t/Foo/Bar/One".  If I'm fixing a bug in
>>> Foo::Bar::One::B,
>>> I run "Build test test_dirs=t/Foo/Bar/One/B/t" over and over, and 
>>> then
>>> when I'm
>>> satisfied I run a full regression test with "Build test".
>>
>> That works fine if you organize tests by directory, but a lot of
>> authors like to group their tests by function in the /t directory -
>
> This may be a habit formed from MakeMaker's limitation on directory
> depth.

Maybe, but it's also something that's not going to go away anytime 
soon.  And I'm not convinced it's a bad habit (especially not for small 
projects).


>> Using regexps on the path to the test script would let you specify a
>> subset of tests to run in all of these scenarios:
>>
>> 	./Build test matching=t/Foo/Bar/One
>> 	./Build test matching=runthrough
>> 	./Build test matching=t/0*
>>
>> Which could be handy.
>
> I presume you mean globs, not regexes?

Nope, I meant regexes, that's just a typo in my example:

	./Build test matching=t/0.*

Here's some better examples:

	# run all tests found that have 'foo' in their names:
	./Build test matching=Foo

	# for David's example, same as this (recursive) glob:
	./Build test test_files=t/Foo/

	# ... but might also include:
	./Build test test_files=t/Bar/Foo/

Us numbered-test freaks could run sets of tests like this:

	# run tests 50-99:
	./Build test matching=[5-9]\\d

And 'matching' can co-exist with globs, which are also a good idea:

	# run all tests in t/Foo ending in "oo.t"
	./Build test test_files=t/Foo/* matching=oo\\.t

I've attached a patch based on your glob one that implements this.

-Steve