882461 - add all of mozbase directories to sys.path instead of really reinventing importing

Status: RESOLVED FIXED

(Testing :: Mozbase, defect)

Description

[Jeff Hammel]

See:
* https://bugzilla.mozilla.org/show_bug.cgi?id=849900
* https://bugzilla.mozilla.org/show_bug.cgi?id=877733#c40
* https://bugzilla.mozilla.org/show_bug.cgi?id=707976

Both runxpcshell.py and automation.py.in have some, ahem,
"interesting" magical import logic:
* http://mxr.mozilla.org/mozilla-central/source/build/automation.py.in#27
* http://mxr.mozilla.org/mozilla-central/source/testing/xpcshell/runxpcshelltests.py#25

And, of course, let us not forget our mozinfo importing hack:
* http://mxr.mozilla.org/mozilla-central/source/testing/xpcshell/Makefile.in
* ...and others?

Until bug 849900 can actually be fixed, we should replace this with
something that adds all the mozbase modules to the sys.path, like :ted
suggests in https://bugzilla.mozilla.org/show_bug.cgi?id=877733#c40 :

for d in os.listdir(mozbase):
    sys.path.append(os.path.join(mozbase, d))

Comment 1

[Andrew Halberstadt [:ahal]]

Attachment: Patch 1.0 - add mozbase to sys.path

Did a quick try push for windows:
https://tbpl.mozilla.org/?tree=Try&rev=67da8ec26bba

Comment 2

[Andrew Halberstadt [:ahal]]

Attachment: Patch 2.0 - actually import modules this time

https://tbpl.mozilla.org/?tree=Try&rev=ec3b4019624f

Might help to actually import the modules afterwards...

Comment 3

[Andrew Halberstadt [:ahal]]

Hmm, trychooser told me that syntax would run tests on win7.. but apparently not:
https://tbpl.mozilla.org/?tree=Try&rev=dfe3b9d8de18

Comment 4

[Jeff Hammel]

Comment on attachment 762275 [details] [diff] [review]
Patch 2.0 - actually import modules this time

lgtm;  because this touches the infamous automation.py.in, a full try run may be prudent.  Also, the mozinfo magic for xpcshell should be cleaned from the makefile, but that can be taken as a follow-up

Comment 5

[Jeff Hammel]

pushed to try, full run: https://tbpl.mozilla.org/?tree=Try&rev=89bd79ae8344 ; perhaps overkill but there ya go.

Any objections to pushing this to m-c if this works?

Comment 6

[Jeff Hammel]

On android 2.2 opt: 

TEST-UNEXPECTED-FAIL | xpccheck | test test_bug608142.js is missing from test manifest /builds/slave/try-and-0000000000000000000000/build/js/xpconnect/tests/unit/xpcshell.ini!
make[6]: *** [libs-xpcshell-tests] Error 1
make[5]: *** [tools] Error 2
make[4]: *** [tools_tier_platform] Error 2
make[3]: *** [tier_platform] Error 2
make[2]: *** [default] Error 2
make[1]: *** [realbuild] Error 2
make: *** [build] Error 2

I didn't read all of https://bugzilla.mozilla.org/show_bug.cgi?id=608142 and am not sure if this is a real problem (seems unrelated to the patch, anyway). Another try run is going on https://bugzilla.mozilla.org/show_bug.cgi?id=877733 so if this recurs, then we can decide what to do (again, repeat, not related to this patch).

Again, I'll land this soon sans objections

Comment 7

[(not currently active) Ted Mielczarek]

FYI that's bug 883918, I think it's just an infra problem.

Comment 8

[Jeff Hammel]

pushed to inbound: https://hg.mozilla.org/integration/mozilla-inbound/rev/bb0567fbb4cc

subsequent bug should be filed for Makefile.in cleanup

Comment 9

[Jeff Hammel]

(In reply to Jeff Hammel [:jhammel] from comment #8)
<snip/>
> subsequent bug should be filed for Makefile.in cleanup

bug 884181

Comment 10

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/bb0567fbb4cc