bpo-38018: increase code coverage for multiprocessing.shared_memory

[vinay0410]

https://bugs.python.org/issue38018

[applio]

The first two checks look like good additions.
The use of mock to control the name of the shared memory segments does not look right to me -- have you verified that this test passes?

[vinay0410]

@applio , I have verified that the test passes. Also, I have verified that it covers these lines.
I wanted to add a test case which verifies the fact that nothing breaks, if _make_filename generates name of already existing segment.
I only could ensure that using mock.

[applio]

The test involving mock appears use two distinct names for the two SharedMemory(create=True, size=1) invocations. Crudely adding a print() to show shm1.name and shm2.name seems to confirm this. Hence the two tests that pass are against names[0] and later names[1].

Perhaps I misunderstood but I thought the intent was to verify that _make_filename creating duplicate names does not trigger a problem in and of itself.

[vinay0410]

@applio, The intent of the test is to verify if a problem is caused when _make_filename returns as existing name.
So, the first call shm1 = shared_memory.SharedMemory(create=True, size=1) will create a shared memory with name test01_fn, since such a name(shared memory) doesn't exists yet.

Then I reset the side effect function of the mock object by mock_make_filename.side_effect = names

So, _make_filename will now again return test01_fn on first call. And, test01_fn is now a name of existing shared memory segment, which was created by the first call.
This will result in a FileExistsError here

And then the loop will continue, calling _make_filename again, which will then return test02_fn, and since no shared memory with such name exists, it will be created.

So, therefore it verifies that existing segment names returned by _make_filename will be skipped, and it will be called again.

[applio]

Ahhh, thank you for the clarification. Got it.

[miss-islington]

Thanks @vinay0410 for the PR, and @applio for merging it 🌮🎉.. I'm working now to backport this PR to: 3.8.
🐍🍒⛏🤖

[miss-islington]

Sorry @vinay0410 and @applio, I had trouble checking out the 3.8 backport branch.
Please backport using cherry_picker on command line.
cherry_picker d14e39c8d9a9b525c7dcd83b2a260e2707fa85c1 3.8

[pablogsal]

This PR introduced a regression in FreeBSD buildbots:

https://buildbot.python.org/all/#/builders/168/builds/1417