727258 - Suite Depend Locales broken after Bug 722262

Status: RESOLVED FIXED

(Firefox Build System :: General, defect)

Description

[Justin Wood (:Callek)]

Attachment: Use MOZ_BUILD_APP to help find real source dir [Empty patch]

So, it seems that suite using the mozilla/build application.ini changed behavior slightly.

It used to contain comm-central as the SOURCE_STAMP there, but now it contains Mozilla's Source Stamp.

I'm attaching a patch that fixes this, by reverting to previous behavior with that ini for us, but without breaking Firefox.

I don't have proof yet but I do believe this breaks more automation for us than just L10n Depend Builds

Comment 1

[Justin Wood (:Callek)]

Attachment: Use MOZ_BUILD_APP to help find real source dir [Checked in: Comment 3]

curtains up this time

Comment 2

[Justin Wood (:Callek)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/74d1974e6934

Comment 3

[Marco Bonardo [:mak]]

https://hg.mozilla.org/mozilla-central/rev/74d1974e6934

Comment 4

[Justin Wood (:Callek)]

Comment on attachment 597189 [details] [diff] [review]
Use MOZ_BUILD_APP to help find real source dir
[Checked in: Comment 3]

Review of attachment 597189 [details] [diff] [review]:
-----------------------------------------------------------------

::: build/Makefile.in
@@ -88,1 @@
>  ifdef MOZ_SOURCE_STAMP

Ugh, MOZ_SOURCE_STAMP ?= here is not good enough, we need :=

Too tired to write the patch now, but happy if someone else does.

Comment 5

[Serge Gautherie (:sgautherie)]

Attachment: (Bv1) build/Makefile.in: force MOZ_SOURCE_STAMP value [Backed out: Comment 7]

Per comment 4.
(Untested.)

Comment 6

[Serge Gautherie (:sgautherie)]

Comment on attachment 597652 [details] [diff] [review]
(Bv1) build/Makefile.in: force MOZ_SOURCE_STAMP value
[Backed out: Comment 7]

https://hg.mozilla.org/mozilla-central/rev/966fb7b0873f

Comment 7

[Phil Ringnalda (:philor)]

Backed out in https://hg.mozilla.org/mozilla-central/rev/a853f4017192

Comment 8

[Serge Gautherie (:sgautherie)]

(In reply to Phil Ringnalda (:philor) from comment #7)
> Backed out in https://hg.mozilla.org/mozilla-central/rev/a853f4017192

Patch Bv1 passed on all desktop platforms,
but broke all test (not build) runs on Android (XUL) opt :-/

Example:
https://tbpl.mozilla.org/php/getParsedLog.php?id=9374674&tree=Firefox&full=1
Android XUL Tegra 250 mozilla-central opt test mochitest-4 on 2012-02-15 19:44:27 PST for push 966fb7b0873f
{
========= Started no change (results: 2, elapsed: 0 secs) ==========
python /builds/tegra-032/test/tools/buildfarm/utils/printbuildrev.py fennec
[...]
Traceback (most recent call last):
  File "/builds/tegra-032/test/tools/buildfarm/utils/printbuildrev.py", line 16, in <module>
    app_revision = appini.get('App', 'SourceStamp')
  File "/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/ConfigParser.py", line 541, in get
    raise NoOptionError(option, section)
ConfigParser.NoOptionError: No option 'sourcestamp' in section: 'App'
program finished with exit code 1
elapsedTime=0.051612

======== Finished no change (results: 2, elapsed: 0 secs) ========
}

Comment 9

[Justin Wood (:Callek)]

(In reply to Serge Gautherie (:sgautherie) from comment #8)
> (In reply to Phil Ringnalda (:philor) from comment #7)
> > Backed out in https://hg.mozilla.org/mozilla-central/rev/a853f4017192
> 
> Patch Bv1 passed on all desktop platforms,
> but broke all test (not build) runs on Android (XUL) opt :-/
> 

Ugh, thanks (this is partly why I didn't want to do the patch last night :-) )

Anyway, from log of *build*

make -C pgo export
/tools/python/bin/python2.5 /builds/slave/m-cen-andrd/build/config/Preprocessor.py -DGRE_MILESTONE=13.0a1 -DAPP_BUILDID=20120215190122 -DMOZ_APP_VERSION="13.0a1" -DMOZ_SOURCE_REPO="http://hg.mozilla.org/mozilla-central" -DMOZ_APP_BASENAME="Fennec" -DMOZ_APP_VENDOR="Mozilla" -DMOZ_APP_ID="{aa3c5121-dab2-40e2-81ca-7ea25febc110}"  -DMOZILLA_OFFICIAL -DMOZ_EXTENSION_MANAGER -DOSTYPE=\"Linux\" -DOSARCH=Linux /builds/slave/m-cen-andrd/build/build/application.ini.in > application.ini

Which is not actually showing a sourcestamp

it turns out that when you use -R to find a hg directory, you must point at the root (where .hg is)

 hg -R mobile/xul parent --template="{node|short}\n"
abort: repository mobile/xul not found!

But if you cd to the directory first, hg properly gives you the output
Justin@ORION /d/sources/comm-central/mozilla/mobile/xul
$ hg  parent --template="{node|short}\n"
74d1974e6934

So we need to wrap this in a cd like the next line. (*and* do the := )

Comment 10

[Justin Wood (:Callek)]

Attachment: take 2 [Checked in: Comment 12]

This patch has r+=khuey over IRC.

He asked me to run it past try, even after testing locally (Bv1 should have been tested before pushing of course :/ )

Anyway, this works, but I'm now too tired to watch the tree, so if someone can push for me great! Otherwise I'll weave it into my work tomorrow.

https://tbpl.mozilla.org/?tree=Try&rev=de058221c837

For Reference, an android build shows MOZ_SOURCE_STAMP properly for m-c here -- I also ran android M1 for verification in the try:

/tools/python/bin/python2.5 /builds/slave/try-andrd-xul/build/config/Preprocessor.py -DGRE_MILESTONE=13.0a1 -DAPP_BUILDID=20120216150055 -DMOZ_APP_VERSION="13.0a1" -DMOZ_SOURCE_STAMP="de058221c837" -DMOZ_SOURCE_REPO="http://hg.mozilla.org/try" -DMOZ_APP_BASENAME="Fennec" -DMOZ_APP_VENDOR="Mozilla" -DMOZ_APP_ID="{a23983c0-fd0e-11dc-95ff-0800200c9a66}"  -DMOZILLA_OFFICIAL -DMOZ_EXTENSION_MANAGER -DOSTYPE=\"Linux\" -DOSARCH=Linux /builds/slave/try-andrd-xul/build/build/application.ini.in > application.ini

Comment 11

[Justin Wood (:Callek)]

for c-n, a direct m-c is much preferred (as it will fix test reporting and running l10n stuff) but an m-i push is acceptable failing your ability/willingness to push directly to m-c.

Comment 12

[Justin Wood (:Callek)]

https://hg.mozilla.org/mozilla-central/rev/2bfd0fe5da97