Access Dispatcher via context so that each test can install own dispatcher.

The flaky failure was because the dispatcher object was a global variable, and
therefore, it was unintentionally shared between tests.

Test in assingee_test.go and issue_test.go didn't even invoke
registerTaskHandlers(), but it was OK, as long as backend_test.go gets executed
before assignee_test.go and issue_test.go. Tests in assignee_test.go and
issue_test.go used accidentally the dispatcher set by backend_test.go.

Interestingly, it seems that go-test runs all the tests within a single process
with the same order repeatedly, at least in my workstation. Thus, repeatedly
executing go-test over 1000 times doesn't produce the error in my workstation,
but does randomly in some other bots at a high rate.

I ran CQ runs multiple times to ensure that the test results are stable.

R: tandrii
Bug: 963695
Change-Id: Id66707d2a03a34746faccf1bfb46263c113c60a1
Reviewed-on: https://chromium-review.googlesource.com/c/infra/infra/+/1614598
Commit-Queue: Scott Lee <ddoman@chromium.org>
Reviewed-by: Andrii Shyshkalov <tandrii@chromium.org>
Cr-Commit-Position: refs/heads/master@{#23023}
6 files changed
tree: 6aa90fb69dd18c186c4dc825adfa56cf5684fadb
  1. 3pp/
  2. appengine/
  3. appengine_module/
  4. bootstrap/
  5. build/
  6. cipd/
  7. crdx/
  8. data/
  9. doc/
  10. docker/
  11. glyco/
  12. go/
  13. infra/
  14. node/
  15. packages/
  16. recipes/
  17. test/
  18. utils/
  19. .gitattributes
  20. .gitignore
  21. codereview.settings
  22. CONTRIBUTING.md
  23. DEPS
  24. LICENSE
  25. navbar.md
  26. OWNERS
  27. PRESUBMIT.py
  28. README.md
  29. run.py
  30. test.py
  31. WATCHLISTS
  32. WHITESPACE
README.md

infra.git repository

Welcome to the Chrome Infra repository!

Wondering where to start? Check out General Chrome Infrastructure documentation. In particular, to check out this repo and the rest of the infrastructure code, follow the instructions here. The rest of this page is specific to this repo.

Entry points

  • run.py: wrapper script to run programs contained in subdirectories without having to deal with sys.path modifications.
  • test.py: multi-purpose script to run tests.
  • packages/infra_libs/: generally useful functions and classes
  • infra/services/: standalone programs intended to be run as daemons.
  • infra/tools: command-line tools, intended to be run by developers.
  • appengine/: many Chrome-infra-managed AppEngine applications
  • infra/experimental: for, well, experimental stuff. Once they are stabilized and reviewed, they should be moved in a more permanent place.

Miscellaneous technical stuff

  • bootstrap/: utilities to set up a proper Python virtual environment.
  • infra/path_hacks: submodules of this modules give access to modules in the build/ repository. from infra.path_hacks.common import <stg> is actually getting <stg> from build/scripts/common.
  • utils/: purpose? utils?
  • Need to bump infra/deployed to pick up changes?
    • git push origin <updated hash>:deployed
    • mail chrome-troopers@, include:
      • previously deployed hash (for quick rollback)
      • the hash you just pushed
      • the list of CLs that made this push necessary
      • the output of the git push command

Integrating tests with test.py

If you've added a new module, integrate your tests with test.py:

  1. Create a .coveragerc file in the root directory of the module you want to test. Take a look at another .coveragerc to see what to include in that.
  2. Create a “test” directory in the root directory of the module you want to test. More your *_test.py files to this directory.

Double-check that your tests are getting picked up when you want them to be: ./test.py test <path-to-package>.

Tests still not getting picked up by test.py? Double-check to make sure you have init.py files in each directory of your module so Python recognizes it as a package.