Bug: `sanitizers_functional_test.sh` influences globally `bazel build //:docs`

[PiotrKorkus]

Description

Repositories that have dependency to score_communication are facing issue when building docs with bazel build //:docs

During execution an error in ide_support is reported:

  File "/home/pko/.cache/bazel/_bazel_pko/3c6e3ebec6d1f13bc4ad8e79f49ea7cf/execroot/_main/bazel-out/k8-opt-exec-ST-d57f47055a04/bin/external/score_process+/sphinx_build.runfiles/score_docs_as_code+/src/extensions/score_plantuml.py", line 64, in get_runfiles_dir
    sys.exit(
SystemExit: Could not find runfiles_dir at /home/pko/.cache/bazel/_bazel_pko/3c6e3ebec6d1f13bc4ad8e79f49ea7cf/execroot/_main/bazel-bin/ide_support.runfiles. Have a look at README.md for instructions on how to build docs.

It turns out bazel cannot find runfiles in expected directories and it happens due to the script in communications:

communication/quality/sanitizer/sanitizers_functional_test.sh

Line 25 in 8c1f29e

export RUNFILES_DIR="$0.runfiles"

In which RUNFILES_DIR is set messing later execution of ide_support.

Commenting out bazel_dep to communication solves the issue and both ide_support and docs pass.

Error Occurrence Rate

None

How to reproduce

e.g.
https://github.com/eclipse-score/persistency/actions/runs/21214394044/job/61151306834?pr=172

Supporting Information

eclipse-score/tooling#103
eclipse-score/docs-as-code#295
eclipse-score/docs-as-code#304

Classification

minor

Affected Version

0.5

Expected Closure Version

0.5

Category

• Safety Related
• Security Related

[castler]

I did not look into this yet further - but there is something really wrong :)

The way how sanitizers_functional_test.sh is setting up runfiles, is the correct and standardized way for shell scripts to work with runfiles. Thus, we are not doing anything wrong. Besides that, as per my knowledge and I just tried it out, the RUNFILES_DIR variable is not leaking out of the shell script when normally executed.

~$ echo $RUNFILES_DIR
~$ bazel test //quality/sanitizer:sanitizers_functional_test
Starting local Bazel server (8.3.0) and connecting to it...
INFO: Analyzed target //quality/sanitizer:sanitizers_functional_test (147 packages loaded, 1119 targets configured).
INFO: Found 1 test target...
Target //quality/sanitizer:sanitizers_functional_test up-to-date:
bazel-bin/quality/sanitizer/sanitizers_functional_test
INFO: Elapsed time: 161.543s, Critical Path: 155.67s
INFO: 6 processes: 4 internal, 2 local.
INFO: Build completed successfully, 6 total actions
//quality/sanitizer:sanitizers_functional_test PASSED in 155.6s
Executed 1 out of 1 test: 1 test passes.
~$ echo $RUNFILES_DIR

Besides that - can you please elaborate, why and how your docs build has a dependency to our infrastructure test? and how it is executing it?

Thanks!

[PiotrKorkus]

I agree that something went wrong and that's why I initially created also eclipse-score/tooling#103 as I am still not sure what the root cause is.

The origin of the problem is that executing bazel run //:docs in logging repo on https://github.com/qorix-group/logging/tree/piotrkorkus_test_docs fails with SystemExit: Could not find runfiles_dir as mentioned. Removing bazel_dep to communication solves the issue and build is successful.

So when I found that issue points to communication repo and there is bash script modifying runfiles variable I simply connected the dots. I still find it hard to debug any bazel failures so I would appreciate if you could help to triage the root cause.

[castler]

I am sorry, but we have no knowledge about your docs build and how you utilize it.

Once this information is provided:

can you please elaborate, why and how your docs build has a dependency to our infrastructure test? and how it is executing it?

we can try to help again.

[PiotrKorkus]

The problem no longer occurs when using docs-as-code v3.0.0 and tooling v1.1.0.
Fixed with reworked logic of runfiles collection in eclipse-score/docs-as-code#378