On Mon, Nov 9, 2009 at 4:52 PM, felix <[email protected]> wrote:

> I like one test per file because it makes it easier to
> isolate one test when trying to work on a problem, ...
>

Well, the main isolation mechanism in that mode is in choosing which test(s)
to *run*, rather than which test(s) to be compiled together.

... and because it bypasses some commit-merge problems.
> Anytime people are appending things to a file
> simultaneously, there's a merge conflict that needs
> to be resolved manually.
>

We merge JUnit files with multiple all the time.


> What problem do you see with a large number of files?
> The gcc testsuite is a large collection of pretty small
> files, many of them only a dozen or so lines.
>

Yeah -- despite what I said above, I have no real objective objection. So
why not? Let's try it.


> Recombining all the single tests into one mega-test
> seems fine for now, but it probably won't scale,
> since domita_test is already close to hitting the
> "this script is taking too long" on some browsers.
>

Ah.


> So I think at some point the tests should be
> split into subdirs, where each subdir would be
> combined into a subdir-mega-test that would be run
> individually by the test driver.
>

That's a good idea.

Ihab

-- 
Ihab A.B. Awad, Palo Alto, CA

Reply via email to