964367 - Capability platformName offered by Marionette is incorrect

Status: RESOLVED FIXED

(Remote Protocol :: Marionette, defect)

Description

[Andreas Tolfsen ❲:ato❳]

The capability platformName that's returned on the creation of a new session in Marionette through newSession is incorrectly formatted.

One of the following values is expected on desktop:

    ANDROID
    IOS
    LINUX
    MAC
    UNIX
    WINDOWS
    ANY

The following is returned on Linux:

    Linux

So for desktop this should be a matter of upper-casing the string.

On Firefox OS we should use "MOZ-FIREFOXOS" as suggested by the WebDriver specification, which could eventually deprecate the "b2g" capability.

Comment 1

[Andreas Tolfsen ❲:ato❳]

Attachment: 0001-Bug-964367-Fix-incorrect-format-on-platformName-capa.patch

try: https://tbpl.mozilla.org/?tree=Try&rev=9df3aef7cc25

Comment 2

[Andreas Tolfsen ❲:ato❳]

Tip of inbound is busted, will wait and trigger a new try run once inbound has been fixed.

Comment 3

[Andreas Tolfsen ❲:ato❳]

New try based on central: https://tbpl.mozilla.org/?tree=Try&rev=e33ab53723ef

Comment 4

[Malini Das [:mdas] - Away, not checking bugmail]

Comment on attachment 8371590 [details] [diff] [review]
0001-Bug-964367-Fix-incorrect-format-on-platformName-capa.patch

Review of attachment 8371590 [details] [diff] [review]:
-----------------------------------------------------------------

::: testing/marionette/marionette-server.js
@@ +574,5 @@
> +      // Proprietary extensions
> +      "XULappId" : Services.appinfo.ID,
> +      "appBuildId" : Services.appinfo.appBuildID,
> +      "device": qemu == "1" ? "qemu" : (!device ? "desktop" : device),
> +      "version": Services.appinfo.version

I like the categorization, it's much easier to read now, thanks!

Comment 5

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/76cb0c973a15

Comment 6

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/76cb0c973a15

Comment 7

[Andreas Tolfsen ❲:ato❳]

Thanks for the pointer, and I'm glad to see jgriffin uplifted it.  It's really difficult to spot these incompatibilities because they don't trigger any test errors in try…