[llvm] 258f055 - [Orc][examples] Add LLJITWithRemoteDebugging example

[David Blaikie via llvm-commits]

On Tue, Apr 6, 2021 at 2:00 AM Stefan Gränitz <stefan.graenitz at gmail.com> wrote:
>
> On 05/04/2021 19:16, Eric Christopher wrote:
>>
>> @Eric Christopher - any idea what the ToolSubst's you updated in https://reviews.llvm.org/rGea7b1c79f73d8def are for/when they're needed/not needed?
>
>
> Sorry, been too long I can't recall :(
>
> Ok, I had a look an explained the situation in https://reviews.llvm.org/D99931
>
> Maybe the test suite could use a warning for RUN lines with relative paths, i.e. without ToolSubst?

Sounds plausible - though I admit to not knowing a great deal about
it. But yeah, if there are cases where the tests could end up running
an arbitrary binary installed on the system, that'd be bad.

> And: Is there a way to query for builders that match a CMake setting?

Not that I know of. Maybe looking in the zorg repo (where the buildbot
config is) it'd be possible to figure out the configuration python and
find them that way?

> I only know these two build with examples enabled:
> http://lab.llvm.org:8011/#/builders/13
> http://lab.llvm.org:8011/#/builders/61
>
>
> On Sat, Apr 3, 2021 at 12:08 PM David Blaikie <dblaikie at gmail.com> wrote:
>>
>> There are bots that build with examples enabled, but they do the same thing I used to do: configure with examples and check-llvm. This way the issue doesn't show up.
>
>
> Hmm, it does seem to show up for me (after I deleted the binary and removed the CMakeFiles.txt change) in check-llvm and check-all. So I'm not sure what's different about your situation or buildbots, etc - do you use Ninja or make or some other thing? (though I'd imagine we have buildbots using ninja... )
>
> Interesting, with a clean build it now shows up here too. Maybe I really didn't clean up the build directory when I ran the test suite earlier. Anyway, yes you are right. It's necessary to add them to LLVM_TEST_DEPENDS manually.
>
>
> On Sat, Apr 3, 2021 at 12:08 PM David Blaikie <dblaikie at gmail.com> wrote:
>
> In which bit of python? In the ToolSubst code? I think the "unresolved=ignore" part means it doesn't need to be exactly the same as the conditions in which its built
>
> Yes sounds reasonable.
>