Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Clean test_{runner,scalafmt_helper}, add diagnostic env vars, and improve test_cross_build performance #1646

Merged
merged 2 commits into from
Nov 12, 2024
Merged

Clean test_{runner,scalafmt_helper}, add diagnostic env vars, and improve test_cross_build performance #1646

merged 2 commits into from
Nov 12, 2024

Commits on Nov 9, 2024

  1. Clean test_{runner,scalafmt_helper}, add env vars 

    Did a little cleaning up while trying to get protobuf to stop
recompiling so often. Didn't impact that problem, but good cleanups
regardless.

Also added two environment variables to make diagnosing test failures
(or slowdowns) easier:

- Defining `RULES_SCALA_TEST_ONLY`will cause only one specific test case
  to run, with full Bazel output. This is helpful for pinpointing and
  working on a specific test without hunting for it or recreating its
  commands.

- Defining `RULES_SCALA_TEST_VERBOSE` will cause all tests to emit full
  Bazel output. This is more useful when running all the tests from a
  specific script from `test/shell/test_*.sh`.

Also, today I learned that `SECONDS` is a special Bash variable. Making
it `local` screwed it up and showed all tests as taking zero seconds.
    @mbland
    mbland committed Nov 9, 2024
    Configuration menu
    Copy the full SHA
    2f6efef View commit details
    Browse the repository at this point in the history

  2. Don't clean test_cross_build, shutdown at end 

    This speeds up the `test_cross_build.sh` tests substantially,
particularly between runs. Here are the times for two runs before this
change; the first is after running `bazel clean` and `bazel shutdown` in
the `test_cross_build` directory:

```txt
$ /usr/bin/time ./test_cross_build.sh

running test test_cross_build
 Test "test_cross_build" successful (73 sec)
running test test_scalafmt
 Test "test_scalafmt" successful (94 sec)
      167.01 real         0.37 user         0.50 sys

$ /usr/bin/time ./test_cross_build.sh

running test test_cross_build
 Test "test_cross_build" successful (8 sec)
running test test_scalafmt
 Test "test_scalafmt" successful (92 sec)
       99.96 real         0.35 user         0.48 sys
```

Here are the times for two equivalent runs after this change:

```txt
$ /usr/bin/time ./test_cross_build.sh

running test test_cross_build
 Test "test_cross_build" successful (71 sec)
running test test_scalafmt
 Test "test_scalafmt" successful (6 sec)
       77.50 real         0.33 user         0.44 sys

$ /usr/bin/time ./test_cross_build.sh

running test test_cross_build
 Test "test_cross_build" successful (5 sec)
running test test_scalafmt
 Test "test_scalafmt" successful (2 sec)
        7.21 real         0.26 user         0.38 sys
```

The effect is even more dramatic when setting compiler flags in
`.bazelrc`, as I've tried while building protobuf-v28.3 under Bazel
6.5.0.
    @mbland
    mbland committed Nov 9, 2024
    Configuration menu
    Copy the full SHA
    4faf196 View commit details
    Browse the repository at this point in the history