targets incompatible with the current platform are selected about target-determinator HOT 5 CLOSED

Thanks for the pointers. I will give it a shot.

from target-determinator.

I think for each affected target bazel cquery --config=asan "config(target, id)" must be called to check if the target should be selected.

https://bazel.build/query/cquery#config

from target-determinator.

Thanks for the bug report! Yeah, this makes sense - I think the right place for us to fix this is in MatchingTargets:

There are two approaches I can see to going about changing it:

1. Either when we're populating it, or when we're returning targets, run a cquery like this to filter to just the targets in the right configuration
2. We do have all of the configuration details in this struct, we just don't use it - if we knew the configuration we wanted to build for, we could do this just with in-memory map lookups without needing to call out to Bazel again

I'd probably lean towards 1 because a clear "filter via an opaque call to bazel" is more obviously correct than "work out the correct configuration ID and filter by it", but I think either could work.

Could I interest you in putting together a PR? I imagine it would take the shape of running one more cquery to filter this slice:

from target-determinator.

Hi Daniel, I looked into your suggestions. I have not yet decided which option to use. But was wondering if the second one is even feasible. Is this information available somewhere - "if we knew the configuration we wanted to build for"? Is there a command to query the configuration id given bazel-opts?

from target-determinator.

Is there a command to query the configuration id given bazel-opts?

After running bazel cquery, if you run bazel config, you'll get a dump of configurations which exist - some will be explicitly labelled as "exec" or "host" configs, and others won't be. I believe any that aren't exec or host, are target configs which we may be attempting to build for. But I'll be honest - I'm not completely sure on this part. It definitely leads me to prefer approach 1, but if someone wants to do the research to work out a viable way to do 2, that could be interesting too :)

from target-determinator.