900830 - dom/plugins/test/unit/test_nice_plugin_name.js fails when using the non-shared plugins directory on Windows

Status: RESOLVED WONTFIX

(Core Graveyard :: Plug-ins, defect)

Description

[Mihnea Dobrescu-Balaur (:mihneadb)]

We introduced a change in the xpcshell test harness that gives every test its own plugins dir to allow running tests concurrently.

This is not really used right now on Windows because of the xpcshell.cpp defaulting to mGREdir/plugins (see bug 899857). I ran the dom/plugins tests on a releng test slave while making sure that they were using the temporary isolated plugins dir, and it turns out that test_nice_plugin_name.js fails.

I suspect it's because of get_test_plugintag (from head_plugins.js), but I don't know all the implementation details.

[I'm referring to the part where the test does:
 * copy nptestplugin to npblah
 * move nptestplugin away to tmpdir
 * check expected perm string ( which yields nptest instead of npblah )
]

It seems that the expected_perm_string looks at the plugin file in the *shared* plugins directory (i.e. mGREdir/plugins) which is untouched, so it has nptest, not npblah.

Comment 1

[Georg Fritzsche [:gfritzsche]]

Do you have a TBPL link handy for the failures?
Note: I may be slow to respond until monday.

Comment 2

[Mihnea Dobrescu-Balaur (:mihneadb)]

This is from the test slave I played with.

http://pastebin.mozilla.org/2763810

BTW.. -- this is from there as well :)

INFO | Result summary:
INFO | Passed: 1798
INFO | Failed: 3
INFO | Todo: 3

Almost there!

Comment 3

[Georg Fritzsche [:gfritzsche]]

(In reply to Mihnea Dobrescu-Balaur (:mihneadb) from comment #0)
> We introduced a change in the xpcshell test harness that gives every test
> its own plugins dir to allow running tests concurrently.
[...]
> It seems that the expected_perm_string looks at the plugin file in the
> *shared* plugins directory (i.e. mGREdir/plugins) which is untouched, so it
> has nptest, not npblah.

So, this is still with both $gredir/plugins and $profile/plugins in the plugins dir list?
If so, this should be resolved by bug 899857 and is not really worth looking into on its own.

Comment 4

[Mihnea Dobrescu-Balaur (:mihneadb)]

So this should be fixed by bug 899857 but I did encounter the failure again on windows once after that landed. I've marked the tests to run-sequentially for now and we'll try going through them later on after this lands, because this particular bug has been a blocker for quite a while.