1078659 - Rewrite bzpost tests to use Dockerized BMO

Status: RESOLVED FIXED

(Developer Services :: Mercurial: bzpost, defect)

Description

[Gregory Szorc [:gps]]

We can now run a BMO instance from Docker during tests. Tests can thus send actual HTTP requests. Let's rewrite bzpost's tests to hit an actual BMO server instead of mocking things.

Comment 1

[Gregory Szorc [:gps]]

Attachment: bzpost: convert tests to docker ()

Now that we can run a local BMO instance via Docker, we convert the
bzpost tests to use it.

This initial patch does the conversion by setting up a BMO instance,
creating empty bugs, and changing the bugs referenced in the existing
tests. This results in hashes changing, which results in more test churn
than would be desirable.

Since we no longer use mocks.py as part of the test, we remove it and
the hack in local-mozilla-repos.py to have all requests-driven HTTP
requests be mocked, which mocks.py was the only consumer of.

Comment 2

[Gregory Szorc [:gps]]

Attachment: bzpost: remove old logging of bug activity ()

As part of 1e900584cb2d, we added logging code to the bzpost extension
to facilitate testing. This was to work around testing deficiency around
use of mocks.

Now that we run a local BMO instance as part of the tests, we can dump
the state of BMO as part of tests. This is a better method of testing
than mocking because we are actually testing that Bugzilla is updated,
not just that we think we're updating Bugzilla. So, we remove the code
for dumping what comments Bugsy is writing and replace it with dumping
of actual Bugzilla state.

Comment 3

[Mike Hommey [:glandium]]

Comment on attachment 8500621 [details] [diff] [review]
bzpost: convert tests to docker ()

Review of attachment 8500621 [details] [diff] [review]:
-----------------------------------------------------------------

LGTM, but I'm not convinced this is something we need done. To me, this only adds a lot of overhead to test bugsy, which should be tested on its own, not through things using it.
I'll also note this: what's the point of getting review on this when there's been no review whatsoever on the related docker stuff that landed.

::: hgext/bzpost/tests/test-post.t
@@ +1,2 @@
>  #require docker
>    $ $TESTDIR/testing/docker-control.py build-bmo

This is not in version-control-tools.

Comment 4

[Mike Hommey [:glandium]]

Comment on attachment 8500622 [details] [diff] [review]
bzpost: remove old logging of bug activity ()

Review of attachment 8500622 [details] [diff] [review]:
-----------------------------------------------------------------

Same remark as for the other patch.

::: hgext/bzpost/tests/test-post.t
@@ +78,5 @@
> +
> +  $ $TESTDIR/testing/bugzilla.py dump-bug 1
> +  Bug 1:
> +    comments:
> +    - {id: 1, text: ''}

Why are there empty comments all around the place?

Comment 5

[Gregory Szorc [:gps]]

The empty initial comments are because the create bug helper doesn't create an initial bug comment by default.

Comment 6

[Gregory Szorc [:gps]]

https://hg.mozilla.org/hgcustom/version-control-tools/rev/41a4ffcd64e0
https://hg.mozilla.org/hgcustom/version-control-tools/rev/28336e80513f