1031390 - Marionette busted on v2.0

Status: RESOLVED FIXED

(Remote Protocol :: Marionette, defect)

Description

[Nivi]

Marionette currently only supports mozrunner 5.37 and latest pypi.python.org/simple/mozrunner’s version is 6.0.

This is causing incompatibility issues and marionette is not being set up at all. As a result, we are unable to run any tests.

Comment 1

[Michael Vines [:m1] [:evilmachines]]

Looks like the v2.0 version of Marionette can't handle mozrunner 6.0.   Locking to mozrunner 5.37 in v2.0 could be one option to fix:


diff --git a/testing/marionette/client/requirements.txt b/testing/marionette/client/requirements.txt
index 8188cf8..e13e6a8 100644
--- a/testing/marionette/client/requirements.txt
+++ b/testing/marionette/client/requirements.txt
@@ -2,7 +2,7 @@ manifestdestiny
 mozhttpd >= 0.5
 mozinfo >= 0.7
 mozprocess >= 0.9
-mozrunner >= 5.15
+mozrunner == 5.37
 mozdevice >= 0.22
 moznetwork >= 0.21
 mozcrash >= 0.5

Comment 2

[Michael Vines [:m1] [:evilmachines]]

Maybe a dup of bug 1028254?

Comment 3

[Dave Hunt [:davehunt] [he/him] ⌚BST]

We've already pinned mozrunner in v1.4, so we should just need to pin the version of mozrunner used in mozilla-aurora:

http://hg.mozilla.org/releases/mozilla-aurora/file/d4dc609bcc8a/testing/marionette/client/requirements.txt#l6

I'm cloning mozilla-aurora now and will prepare a patch.

Comment 4

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Note, for running against v1.4 you should use https://pypi.python.org/pypi/marionette_client-mozilla-b2g30_v1_4

When I first read this I thought it was about builds failing in TBPL but now I'm not so sure. If we fix this in mozilla-aurora then we'd need another client in PyPI for running against this branch. I was hoping to avoid that (and wait for a 2.0 release branch).

The other option is to install mozrunner 5.37.1 before installing the marionette client, in which case the dependency is already met and 6.0 will not be installed.

Jonathan: We should still pin mozrunner in mozilla-aurora, but do you think we should create another PyPI package?

Nivi: Does installing mozrunner 5.37.1 first fix the issue for you?

Comment 5

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Attachment: Pin mozrunner to latest from branch. v1.0

Here's a patch to pin the mozrunner version. Another option is to uplift the mozrunner changes to mozilla-aurora, if it's safe to do that.

Comment 6

[Jonathan Griffin (:jgriffin)]

Comment on attachment 8447882 [details] [diff] [review]
Pin mozrunner to latest from branch. v1.0

Review of attachment 8447882 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks.

Comment 7

[Jonathan Griffin (:jgriffin)]

Let's not uplift the mozrunner changes to mozilla-aurora, that's a higher-risk option.

Comment 8

[Nivi]

Yes installing mozrunner 5.37 fixes the issue. We have a temporary fix for v2.0. We hope you guys can fix this so we can get rid of our workaround.

Thanks,
Nivi.


(In reply to Dave Hunt (:davehunt) from comment #4)
> Note, for running against v1.4 you should use
> https://pypi.python.org/pypi/marionette_client-mozilla-b2g30_v1_4
> 
> When I first read this I thought it was about builds failing in TBPL but now
> I'm not so sure. If we fix this in mozilla-aurora then we'd need another
> client in PyPI for running against this branch. I was hoping to avoid that
> (and wait for a 2.0 release branch).
> 
> The other option is to install mozrunner 5.37.1 before installing the
> marionette client, in which case the dependency is already met and 6.0 will
> not be installed.
> 
> Jonathan: We should still pin mozrunner in mozilla-aurora, but do you think
> we should create another PyPI package?
> 
> Nivi: Does installing mozrunner 5.37.1 first fix the issue for you?

Comment 9

[Jonathan Griffin (:jgriffin)]

https://hg.mozilla.org/releases/mozilla-aurora/rev/48e57cf7505f

Comment 10

[Jonathan Griffin (:jgriffin)]

Nivi, can you confirm this is fixed now?

Comment 11

[Nivi]

Hi Jonathan,

Thank you so much for the fix. It does work now.

Thanks,
Nivi.