[Openmp-commits] [PATCH] D78566: [OpenMP] Add scaffolding for negative runtime tests

[Joel E. Denny via Phabricator via Openmp-commits]

jdenny added a comment.

In D78566#1995119 <https://reviews.llvm.org/D78566#1995119>, @jdenny wrote:

> In D78566#1994835 <https://reviews.llvm.org/D78566#1994835>, @Hahnfeld wrote:
>
> > LLVM does some funny `\| \bnot\b` regex (which I have yet to understand), so I leave it up to you if you want to dig into that.
>
>
> I don't understand that either.  That makes sense only when `not` stdin is a pipe.  Grepping for `RUN: not` shows that's not always true.  That's probably why `not` is also in `PATH`, but that makes the substitution seem redundant.  When I remove the substitution (from `lit/llvm/config.py`), check-llvm behaves.  I haven't yet run check-clang, etc.


check-clang also behaves.

>> Alternatively, we can either provide `%not` (which makes it unambiguous)
> 
> My only argument against that solution is that no other LLVM test suite does this.  I grepped for `RUN:.*%not` to verify.
> 
> `\bnot\b` works in a quick example I just tried.
> 
>> or just not wire it up for the other suites and adapt the commit message.
> 
> Maybe.

Hopefully by providing the familiar `not` now, we'll encourage more negative tests.  To that end, I went with the `\bnot\b` substitution.

I checked that it actually works this time by adding a `RUN: not false` in each openmp test suite.  Of course, I didn't include that in the commit.

I'll wait a bit longer to push in case people have comments for this update.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D78566/new/

https://reviews.llvm.org/D78566