how do I run parallel mode via API?

[nedbat]

Originally reported by Anonymous

---

My flow goes like this:

  cov = coverage.coverage(data_suffix=True)
  cov.start()
  suite = unittest.defaultTestLoader.discover(...)
  suite = ConcurrentTestSuite(suite, fork_for_tests(4))  # os.fork 4 subprocesses
  unittest.TextTestRunner().run(suite)
  cov.stop()
  cov.combine()
  cov.save()

My suite does a lot of forks (via http://coreygoldberg.blogspot.com/2013/06/python-concurrencytest-running.html).

I thought setting data_suffix=True would force the parallel runs, yet I get different coverage results depending if I os.fork the runs or not.

---

• Bitbucket: https://bitbucket.org/ned/coveragepy/issue/312
• This issue had attachments: concurrent-example.tgz. See the original issue for details.

[nedbat]

Can you provide a small sample project that demonstrates the problem? (Even a large project that demonstrates the problem would be good.)

[nedbat]

Original comment by Albert Ting (Bitbucket: altlist, GitHub: Unknown)

---

Attached is an example.

#!bash

$ run_tests --parallel=1    -> 100% coverage
$ run_tests --parallel=10  ->  35% coverage

[nedbat]

Original comment by Albert Ting (Bitbucket: altlist, GitHub: Unknown)

---

Hi, are there any updates?

[nedbat]

Thanks, I've reproduced it, and will dig in soon to understand what's going wrong.

[nedbat]

@altlist This seems to be a duplicate of #310, os.fork and os._exit.

[nedbat]

Duplicate of #310.