Bazel 0.25 breaks upb build: Output artifact '_objs/upb_lib/descriptor.upb.pic.o' not under package directory 'external/com_google_protobuf' for target '@com_google_protobuf//:descriptor_proto'

[haberman]

Description of the problem / feature request:

Bazel 0.25 breaks the build of https://github.com/protocolbuffers/upb.

Bugs: what's the simplest, easiest way to reproduce this bug? Please provide a minimal example if possible.

$ wget https://github.com/bazelbuild/bazel/releases/download/0.25.1/bazel-0.25.1-linux-x86_64
$ chmod a+x bazel-0.25.1-linux-x86_64
$ git clone --single-branch --branch bazel25 https://github.com/haberman/upb.git
$ cd upb
$ ../bazel-0.25.1-linux-x86_64 build :descriptor_upbproto

I get the following output/error:

ERROR: /usr/local/google/home/haberman/.cache/bazel/_bazel_haberman/3d6dddcf72cff0e2141218f5c0401c8f/external/com_google_protobuf/BUILD:284:2: in //:build_defs.bzl%_upb_proto_library_aspect aspect on proto_library rule @com_google_protobuf//:descriptor_proto:
Traceback (most recent call last):
        File "/usr/local/google/home/haberman/.cache/bazel/_bazel_haberman/3d6dddcf72cff0e2141218f5c0401c8f/external/com_google_protobuf/BUILD", line 284                          
                //:build_defs.bzl%_upb_proto_library_aspect(...)
        File "/tmp/foo/upb/build_defs.bzl", line 357, in _upb_proto_aspect_impl
                cc_library_func(ctx = ctx, hdrs = files.hdrs, srcs =..., ...]))
        File "/tmp/foo/upb/build_defs.bzl", line 288, in cc_library_func
                cc_common.compile(actions = ctx.actions, feature_con..., <5 more arguments>)                                                                                       
Output artifact '_objs/upb_lib/descriptor.upb.pic.o' not under package directory 'external/com_google_protobuf' for target '@com_google_protobuf//:descriptor_proto'               
ERROR: Analysis of target '//:descriptor_upbproto' failed; build aborted: Analysis of target '@com_google_protobuf//:descriptor_proto' failed; build aborted    

What operating system are you running Bazel on?

Linux.

What's the output of bazel info release?

release 0.25.1

Any other information, logs, or outputs that you want to share?

This seems related to my StackOverflow question where I am wondering how to properly write rules/aspects that operate on external repositories: https://stackoverflow.com/questions/55796252/how-to-write-bazel-rules-that-work-with-external-repositories

I am very receptive to hearing about how to write my rule/aspect to make it simpler and avoid this error.

[haberman]

Any idea @hlopko @oquenchil ?

[oquenchil]

The problem is that here we are calling getPackageName() instead of getPackageIdentifier(), so the repo section of the package got dropped. This problems exists at head. I will send a fix. I can't think of a workaround at the moment.

0.26 is a week away, would it be possible for you to wait until then?

[aehlig]

[...] This problems exists at head. I will send a fix. I can't think of a workaround at the moment.

0.26 is a week away, would it be possible for you to wait until then?

Note that the baseline for 0.26 (#7499) is already determined. Are your requesting a cherry-pick? That would require a rgression with respect to 0.25. Is a patch-release needed for 0.25 (#7498)?

[oquenchil]

I have sent the fix for review. It would need to be cherry-picked into 0.26. I don't know if 0.25 needs to be patched. I would imagine that this depends on how many people this affects. Joshua if I remember correctly from our chat conversations you were still playing around experimenting with this API. Did this make it into any of your releases and does it affect a lot of users?

[haberman]

Did this make it into any of your releases and does it affect a lot of users?

gRPC is an important customer of mine who is more or less blocked until I can get this working. That's why I have been spending so much effort on it. I found a way to get it working finally in 0.24.1 but then 0.25 regressed me.

If 0.26 is a week away, I am ok waiting for that. But I don't want to wait much longer than that.

[oquenchil]

Thanks Joshua. The fix 942f7cf has been cherrypicked already.

[dkelmer]

@haberman I'm also creating a patch release with the fix for this so you'll be able to use 0.25.2 later today

[haberman]

Thanks all for the quick resolution!

[oquenchil]

Is everything working correctly Joshua?

[haberman]

@oquenchil yes I have everything working now, thanks! :)