Remove processArchitecture value from manifests.

Remove processArchitecture value from manifests.


BUG=166496
TEST=Install component build of Chrome via "setup.exe --chrome --multi-install" and make sure you can launch chrome.exe.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=182740

[gab, 2013-02-14 21:01:43]

Follow-up to http://crrev.com/180801 which broke the component install for x86.

The manifests don't even need to specify "processArchitecture" (it seems to be
determined directly from the binary by Windows anyways).

Most of them are likely copies of the first one anyways and don't have this for
a specific reason...

It's annoying that we have to change these all over the place (ref: cpu's
"grrrr" on https://codereview.chromium.org/12222002); the manifests should
mostly all come from the same manifest. I have a change to do this at
https://codereview.chromium.org/10901011/, but it was reverted because the
Sconscript used by the official build is stupid -- I'd have to change the
Sconscript and haven't gone back to this yet.

Thanks,
Gab

[gab, 2013-02-14 21:11:25]

https://codereview.chromium.org/12252043/diff/4012/chrome/tools/build/win/cre...
File chrome/tools/build/win/create_installer_archive.py (right):

https://codereview.chromium.org/12252043/diff/4012/chrome/tools/build/win/cre...
chrome/tools/build/win/create_installer_archive.py:455: "version='0.0.0.0'
language='*'/>"
Note that the modifications in this file alone are sufficient to fix the
component install, but might as well get rid of all of them if they're not
useful... objections?

[gab, 2013-02-14 21:12:49]

The try servers are likely choking on the Windows endlines... don't pay
attention to them...

[gab, 2013-02-14 21:18:06]

Actually, perhaps it makes sense to keep it for
"Microsoft.Windows.Common-Controls"...?! I don't know... thoughts? Try bots are
useless since they can't apply the patch so only way to find out is the
waterfall I guess (works fine locally)...

[jschuh, 2013-02-14 21:18:27]

When I tried removing processorArchitecture the Win64 build failed to launch.
Have you tried building and running chrome.exe with target_arch=x64 ?

[gab, 2013-02-14 22:38:15]

[cpu to CC, since this no longer touches chrome/app]

After discussion with Justin; the changes to the other manifests on the
"Microsoft.Windows.Common-Controls" sections were breaking the x64 build. The
changes on create_installer_archive.py are sufficient to fix the current issue
with the component x86 build.

We can revisit if it doesn't work when the x64 build has a component installer
to try this on.

Cheers!
Gab

[grt (UTC plus 2), 2013-02-15 01:45:06]

lgtm

[gab, 2013-02-15 03:27:28]

FWIW, I thought about this some more; here is how I think manifests work as far
as processArchitecture is concerned (pure speculation, but I think that makes
sense...):

1) The target DLL's manifest can specify a "processArchitecture" or not (it
can't specify "*", or if it does that's just the name of its architecture, not
actually a wildcard).

2) An executable's manifest can specify the name of other manifests as
dependencies (as long as the said dependencies can be found in that executable's
config e.g. chrome.exe.config IIRC); when specifying dependencies: the
executable can say whether it wants to depend on a specific architecture (if it
does specify one, it must match one of the manifests in its config path -- it
can specify "*" in which case it matches any architecture that are compatible
with itself). It can also not specify any architecture (I'm not sure if that
acts like "*", but at least I know it matches with a target manifest that also
doesn't specify any architecture (as in this CL).

tl;dr; "*" in the "target" manifest doesn't make sense, "*" in the executable;s
manifest means it can match either architecture available, as fits it's binary
configuration (i.e., the "*" is converted at runtime by Windows according to the
binary).

Again, speculation, but I thought I'd drop my thoughts here for future reference
(at least I think those assumptions aren't broken by anything I've seen so far
with manifests!).

Cheers!
Gab

[commit-bot: I haz the power, 2013-02-15 03:29:46]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/gab@chromium.org/12252043/12001

[commit-bot: I haz the power, 2013-02-15 16:43:16]

Retried try job too often on win7_aura for step(s) browser_tests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=win7_aura&...

[commit-bot: I haz the power, 2013-02-15 17:00:41]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/gab@chromium.org/12252043/12001

[commit-bot: I haz the power, 2013-02-15 17:24:59]

Change committed as 182740