[llvm] r335202 - [RISC-V] Fix a test case to not include label names as those aren't

Chandler Carruth via llvm-commits llvm-commits at lists.llvm.org
Thu Jun 21 14:00:14 PDT 2018


On Thu, Jun 21, 2018 at 7:50 AM Sanjay Patel via llvm-commits <
llvm-commits at lists.llvm.org> wrote:

>
>
> On Thu, Jun 21, 2018 at 6:55 AM, Alex Bradbury <asb at asbradbury.org> wrote:
>
>> On 21 June 2018 at 06:42, Chandler Carruth via llvm-commits
>> <llvm-commits at lists.llvm.org> wrote:
>> > Author: chandlerc
>> > Date: Wed Jun 20 22:42:05 2018
>> > New Revision: 335202
>> >
>> > URL: http://llvm.org/viewvc/llvm-project?rev=335202&view=rev
>> > Log:
>> > [RISC-V] Fix a test case to not include label names as those aren't
>> > stable in non-asserts builds. This fixes a test failure in release
>> > config.
>>
>> Thanks for the fix, the failure was triggered by a change in label
>> names in r335183.
>>
>> CHECK lines including label names are frequently generated by
>> update_llc_test_checks.py. It looks like .L* # %somename' is checked
>> in 157 files within test/CodeGen/*.
>>
>> $ git grep -l -i "\.L.*# %" test/CodeGen | wc -l
>> 157
>>
>> These don't currently pose a problem for running the test suite in
>> release builds, but arguably it makes the tests more brittle than
>> necessary. Should we be stripping these comments in
>> update_llc_test_checks.py?
>>
>
> Sounds like a good idea (or at least as an option). Adding people who have
> worked on the scripts more recently than me.
>

I think the root cause is that these labels are not stable for this
*particular* test and/or this *particular* target.

We've had no issues using the label names on x86 that I can remember.

I'd suggest keeping the ability to check label names in the tests because
it makes them dramatically more readable, and instead figuring out how to
stabilize the names used in this *particular* test.

(My impression is that these names are actually coming from instructions
not IR label names, because they didn't go away in the opt build they just
changed value. Sorry for taking a quick fix approach, but I just wanted
check-tests to pass. ;])


>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180621/c01fffa8/attachment.html>


More information about the llvm-commits mailing list