Refactor manifests to allow all executables to use the same base manifests as AdditionalManifestFil…

Refactor manifests to allow all executables to use the same base manifests as AdditionalManifestFiles

Introduce additional.manifest with app compatibility guids and UAC asInvoker property (this is the same as the old chrome.exe.manifest, but without the non-common parts, see chrome.manifest below).

Introduce chrome.manifest to keep a property that was set in the previous chrome.(exe|dll).manifest to declare usage of ComCtl32.dll version 6 as is required per http://msdn.microsoft.com/en-us/library/windows/desktop/bb773175(v=vs.85).aspx to get visual styles.

BUG=127765
(not really fixing it, but a follow up to the CL for it)
TEST=Manifests embedded in chrome.exe, chrome.dll, setup.exe, mini_installer.exe in a VS2010 static build are the exact same as before.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=153268

[gab, 2012-08-22 21:21:49]

Following up on http://codereview.chromium.org/10837344/.

I'm not sure whether my `git mv` of chrome.exe.manifest to additional.manifest
(and then modifying the file) will be linked to chrome.exe.manifest in history
so... Do we care/want it to be? If so I think I might have to rename the file in
svn and then do the rest of my changes.

Cheers,
Gab

[robertshield, 2012-08-22 21:43:38]

lgtm, re. the git question, ought to work, but check with maruel.

[gab, 2012-08-23 04:41:21]

@maruel: See concern in first comment.

[M-A Ruel, 2012-08-23 09:45:39]

On 2012/08/23 04:41:21, gab wrote:
> @maruel: See concern in first comment.

http://codereview.chromium.org/download/issue10879020_1_6.diff
vs
http://codereview.chromium.org/download/issue10879020_1_5.diff

The file is more than 50% different so it's not linked. You can hack your
upload.py accordingly and upload again.

[gab, 2012-08-23 20:56:31]

cpu: chrome\app OWNERS approval
thakis: chrome\ OWNERS approval

Thanks,
Gab

PS: It looks like we won't be able to keep the history linked from the previous
manifests (http://crbug.com/144408), but they didn't have much history since
their creation anyways...

[Nico, 2012-08-23 21:19:38]

I know next to nothing about manifests. If robertshield and cpu are both happy,
I'll gladly stamp this though.

[cpu_(ooo_6.6-7.5), 2012-08-23 23:21:29]

the change itself is lgtm but let me tell you an issue we discovered when making
winja embed the manifests

The microsoft design is that there can be N manifests specified in M object
files plus stand-alone ones

so you have : cc -->  link --> exe
                           --> manifest

Then   exe      -->  link --> exe 
       manifest -->

The final step is really bad for ninja, since the input and output are the same
thing

I don't know if this CL worsens that situation or if it is already bad. Please
ask scottmg or nico

[scottmg, 2012-08-23 23:28:33]

On 2012/08/23 23:21:29, cpu wrote:
> the change itself is lgtm but let me tell you an issue we discovered when
making
> winja embed the manifests
> 
> The microsoft design is that there can be N manifests specified in M object
> files plus stand-alone ones
> 
> so you have : cc -->  link --> exe
>                            --> manifest
> 
> Then   exe      -->  link --> exe 
>        manifest -->
> 
> The final step is really bad for ninja, since the input and output are the
same
> thing
> 
> I don't know if this CL worsens that situation or if it is already bad. Please
> ask scottmg or nico

Yeah, I don't think this will work on ninja because it's relying on the magic
handling of .manifest in 'sources' that VS does. The ninja generator on Windows
doesn't emulate that.

I don't know if anyone cares though? I don't need a manifest personally, but grt
and cpu seemed happy that they started working, so might be sad if they stopped
working.

[grt (UTC plus 2), 2012-08-24 01:37:50]

I volunteer the results of my manifest research a few weeks ago:

https://docs.google.com/a/chromium.org/document/d/1Ag4V4kjREQOum1CjivTIGwBVnd...

Winja doesn't do all of that monkeying around.  Life might be easier if we
stopped using any of the VS behavior and just slammed in a complete manifest of
our own creation.  I haven't confirmed that this makes sense, though, so I may
be smokin' the stuff.

[gab, 2012-08-24 03:18:29]

FWIW, this change doesn't affect the current state of things (i.e. the manifests
were sources and still are; although they weren't explicitly stated as such and
thus wouldn't force rebuilds when modified so I added those specific
dependencies as part of this CL to help), the core of this CL though is that now
instead of having separate manifests with the same properties, there is a single
manifest which all exes import. This is better from a code design pov, but
doesn't solve/worsen the currently existing issue with winja.

On 2012/08/24 01:37:50, grt wrote:
> I volunteer the results of my manifest research a few weeks ago:
> 
>
https://docs.google.com/a/chromium.org/document/d/1Ag4V4kjREQOum1CjivTIGwBVnd...
> 
> Winja doesn't do all of that monkeying around.  Life might be easier if we
> stopped using any of the VS behavior and just slammed in a complete manifest
of
> our own creation.  I haven't confirmed that this makes sense, though, so I may
> be smokin' the stuff.

[gab, 2012-08-24 03:24:31]

Also, AdditionalManifestFiles do work in winja AFAIK, they don't get embedded,
but they get generated and put side-by-side which is sufficient for us
developers who work on the installer and care about this.

Nico, lgty?

On 2012/08/24 03:18:29, gab wrote:
> FWIW, this change doesn't affect the current state of things (i.e. the
manifests
> were sources and still are; although they weren't explicitly stated as such
and
> thus wouldn't force rebuilds when modified so I added those specific
> dependencies as part of this CL to help), the core of this CL though is that
now
> instead of having separate manifests with the same properties, there is a
single
> manifest which all exes import. This is better from a code design pov, but
> doesn't solve/worsen the currently existing issue with winja.
> 
> On 2012/08/24 01:37:50, grt wrote:
> > I volunteer the results of my manifest research a few weeks ago:
> > 
> >
>
https://docs.google.com/a/chromium.org/document/d/1Ag4V4kjREQOum1CjivTIGwBVnd...
> > 
> > Winja doesn't do all of that monkeying around.  Life might be easier if we
> > stopped using any of the VS behavior and just slammed in a complete manifest
> of
> > our own creation.  I haven't confirmed that this makes sense, though, so I
may
> > be smokin' the stuff.

[gab, 2012-08-24 19:43:14]

thakis: ping

Looks like your email LGTM reply didn't make it to the code review....

Thanks,
Gab

[scottmg, 2012-08-24 19:49:24]

rubberstamp lgtm, Nico did by mai.

[gab, 2012-08-24 19:51:10]

Thanks Scott, but I need chrome/ OWNER approval. I'll see if I can ping one
live.

[brettw, 2012-08-24 19:52:42]

LGTM rubberstamp