[PATCH] D45215: RFC/WIP: Have lit run the lldb test suite

Zachary Turner via llvm-commits llvm-commits at lists.llvm.org
Wed Apr 4 09:07:40 PDT 2018


Sure, but getting lit to run one file at a time is a nice incremental step
towards that and can make both patches easier to review.
On Wed, Apr 4, 2018 at 9:02 AM Adrian Prantl <aprantl at apple.com> wrote:

>
> On Apr 4, 2018, at 8:53 AM, Pavel Labath <labath at google.com> wrote:
>
>
>
> On Wed, 4 Apr 2018 at 16:47, Zachary Turner <zturner at google.com> wrote:
>
>>
>> On Wed, Apr 4, 2018 at 8:11 AM Jonas Devlieghere via Phabricator <
>> reviews at reviews.llvm.org> wrote:
>>
>>> JDevlieghere added a comment.
>>>
>>> In https://reviews.llvm.org/D45215#1056917, @zturner wrote:
>>>
>>> > I haven’t had time to look at this in detail yet, but when I
>>> originally had
>>> >  this idea I thought we would use lit’s discovery mechanism to find
>>> all .py
>>> >  files, and then invoke them using dotest.py in single process mode
>>> with a
>>> >  path to a specific file.
>>>
>>>
>>> Assuming we can work around the problem of not every `.py` file being a
>>> test (by filtering the `Test` prefix), would there be a way to
>>> differentiate the different test within a single file?
>>>
>> Would we need to?  dotest will just run all the tests in a single file.
>>
>> I can see how it might be desirable as an end state, but not necessarily
>> as an incremental step.
>>
>
> I think I would be fine with not having test-function level resolution in
> v1 of the feature *if* there is a reasonable path forward to make that
> happen in the future.
>
>
> I forgot about the multiple tests-per-py-file scenario. I think it is
> important to completely support this in the LIT driver. The whole point of
> this exercise is to get dotest out of the business of scheduling tests so
> the end-goal should be to have LIT recognize the individual functions as
> tests.
>
> Side-question to Jonas: was your idea to run lit for each variant
> (dwarf,dwo,dsym,...) or to have dotest spawn multiple versions of each test?
>
> The RUN lines proposal seemed to make that hard if not impossible, but
> with a custom test format it seems plausible to reach that state
> incrementally.
>
>
> As an intermediate step, we could also have one RUN line per test
> function, but I see that that is a lot of work to maintain.
>
> -- adrian
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180404/fb7339a8/attachment.html>


More information about the llvm-commits mailing list