Build a Syzygy-reordered mini_installer.exe.

Build a Syzygy-reordered mini_installer.exe.

This CL adds an All_syzygy target to the build/all.gyp solution,
that builds a Syzygy-reordered chrome.dll and mini_installer.
If there there is a chrome.dll-order.json file present in the output
directory, chrome.dll will be ordered according to that, otherwise
it will be randomized.
Moved most of mini_installer.gyp to a common .gypi file for reuse.


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=111710

[robertshield, 2011-11-28 01:16:38]

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
File chrome/installer/mini_installer.gypi (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
chrome/installer/mini_installer.gypi:1: {
license header needed.

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
chrome/installer/mini_installer.gypi:65: 'AdditionalDependencies': [ # Must
explicitly link in VC2005E
two spaces before comment

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
File chrome/installer/mini_installer_syzygy.gyp (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
chrome/installer/mini_installer_syzygy.gyp:1: {
license header

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
chrome/installer/mini_installer_syzygy.gyp:3: 'version_py':
'../../chrome/tools/build/version.py',
I believe there's a <(DEPTH) built-in you can use here and below to avoid
hard-coding directory depth.

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
chrome/installer/mini_installer_syzygy.gyp:11:
'../../build/win_precompile.gypi',
DEPTH (see above)

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
chrome/installer/mini_installer_syzygy.gyp:36: }, { # else branding!="Chrome"
two spaces before comment

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/creat...
File chrome/tools/build/win/create_installer_archive.py (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/creat...
chrome/tools/build/win/create_installer_archive.py:119: if not
os.path.exists(dst_file):
does this mean that the build will not do what is expected if someone has a
stale staging dir left behind?

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/syzyg...
File chrome/tools/build/win/syzygy_reorder.py (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/syzyg...
chrome/tools/build/win/syzygy_reorder.py:16:
os.path.join(os.path.dirname(__file__), '../../../..'))
seems like this should be an input option

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/syzyg...
chrome/tools/build/win/syzygy_reorder.py:23: _BINARIES = [('chrome.dll',
'chrome_dll.pdb'),]
feels like this should be specified in a .gyp file as well. Paths set in helper
scripts are fragile and a bit more irksome to track down.

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/syzyg...
chrome/tools/build/win/syzygy_reorder.py:28: _RANDOM_SEED = 1347344
this is not a comment, but http://xkcd.com/221/

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/syzyg...
chrome/tools/build/win/syzygy_reorder.py:64: # No ordering file, we randomize
the output.
Is the reordered path soon expected to be the default? If so, should this cause
a build warning, or somehow be made very noticeable in the build output when it
takes the random reordering path?

[Sigurður Ásgeirsson, 2011-11-28 15:02:09]

Thanks, please take another look.

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
File chrome/installer/mini_installer.gypi (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
chrome/installer/mini_installer.gypi:1: {
On 2011/11/28 01:16:38, robertshield wrote:
> license header needed.

Done.

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
chrome/installer/mini_installer.gypi:65: 'AdditionalDependencies': [ # Must
explicitly link in VC2005E
On 2011/11/28 01:16:38, robertshield wrote:
> two spaces before comment

Done.

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
File chrome/installer/mini_installer_syzygy.gyp (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
chrome/installer/mini_installer_syzygy.gyp:1: {
On 2011/11/28 01:16:38, robertshield wrote:
> license header

Done.

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
chrome/installer/mini_installer_syzygy.gyp:11:
'../../build/win_precompile.gypi',
On 2011/11/28 01:16:38, robertshield wrote:
> DEPTH (see above)

Apparently DEPTH isn't expanded - or defined - soon enough for this to work for
includes.

http://codereview.chromium.org/8698009/diff/2001/chrome/installer/mini_instal...
chrome/installer/mini_installer_syzygy.gyp:36: }, { # else branding!="Chrome"
On 2011/11/28 01:16:38, robertshield wrote:
> two spaces before comment

Done.

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/creat...
File chrome/tools/build/win/create_installer_archive.py (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/creat...
chrome/tools/build/win/create_installer_archive.py:119: if not
os.path.exists(dst_file):
On 2011/11/28 01:16:38, robertshield wrote:
> does this mean that the build will not do what is expected if someone has a
> stale staging dir left behind?

The script starts by wiping out the staging dir. The same issue would otherwise
occur if a file is removed from the build or release script, but was present in
a staging dir.

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/syzyg...
File chrome/tools/build/win/syzygy_reorder.py (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/syzyg...
chrome/tools/build/win/syzygy_reorder.py:16:
os.path.join(os.path.dirname(__file__), '../../../..'))
On 2011/11/28 01:16:38, robertshield wrote:
> seems like this should be an input option

It's only used to calculate the path to the default tool to use, though.

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/syzyg...
chrome/tools/build/win/syzygy_reorder.py:23: _BINARIES = [('chrome.dll',
'chrome_dll.pdb'),]
On 2011/11/28 01:16:38, robertshield wrote:
> feels like this should be specified in a .gyp file as well. Paths set in
helper
> scripts are fragile and a bit more irksome to track down. 

This is true. However, I was modelling this from the installer archive creation
as I expect we'll optimize more of chrome's exes over time, and it seems like
there's an attempt there to reduce the number of build targets. 
Doesn't make a difference for this iteration, so will change.

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/syzyg...
chrome/tools/build/win/syzygy_reorder.py:28: _RANDOM_SEED = 1347344
On 2011/11/28 01:16:38, robertshield wrote:
> this is not a comment, but http://xkcd.com/221/

:)

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/syzyg...
chrome/tools/build/win/syzygy_reorder.py:64: # No ordering file, we randomize
the output.
On 2011/11/28 01:16:38, robertshield wrote:
> Is the reordered path soon expected to be the default? If so, should this
cause
> a build warning, or somehow be made very noticeable in the build output when
it
> takes the random reordering path?

Good question - I figure once we turn this on for the official builds, we'll
want the build to fail if the ordering file is missing. There'll be a followup
CL once I figure out how best to do that.

[Roger McFarlane (Chromium), 2011-11-28 15:09:27]

http://codereview.chromium.org/8698009/diff/2001/chrome/chrome_syzygy.gyp
File chrome/chrome_syzygy.gyp (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/chrome_syzygy.gyp#new...
chrome/chrome_syzygy.gyp:6: ['OS=="win" & and fastbuild==0', {
& ?

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/creat...
File chrome/tools/build/win/create_installer_archive.py (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/creat...
chrome/tools/build/win/create_installer_archive.py:401: print sys.argv
a remnant debugging aid?

[robertshield, 2011-11-28 15:11:41]

lgtm

[Roger McFarlane (Chromium), 2011-11-28 15:17:28]

lgtm

http://codereview.chromium.org/8698009/diff/8001/chrome/tools/build/win/syzyg...
File chrome/tools/build/win/syzygy_reorder.py (right):

http://codereview.chromium.org/8698009/diff/8001/chrome/tools/build/win/syzyg...
chrome/tools/build/win/syzygy_reorder.py:38: def _ReorderBinary(relink_exe,
executable, symbol, destination_dir):
nit: there's inconsistency between executable and input/output-dll

[Sigurður Ásgeirsson, 2011-11-28 15:24:21]

thanks, committing.

http://codereview.chromium.org/8698009/diff/8001/chrome/tools/build/win/syzyg...
File chrome/tools/build/win/syzygy_reorder.py (right):

http://codereview.chromium.org/8698009/diff/8001/chrome/tools/build/win/syzyg...
chrome/tools/build/win/syzygy_reorder.py:38: def _ReorderBinary(relink_exe,
executable, symbol, destination_dir):
On 2011/11/28 15:17:28, Roger McFarlane (Chromium) wrote:
> nit: there's inconsistency between executable and input/output-dll

Yeah, we should fix the argument name in the relink.exe binary.

[commit-bot: I haz the power, 2011-11-28 15:25:23]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/siggi@chromium.org/8698009/8001

[commit-bot: I haz the power, 2011-11-28 16:47:44]

Change committed as 111710

[Sigurður Ásgeirsson, 2011-11-28 18:55:36]

Flushing dangling comments...

http://codereview.chromium.org/8698009/diff/2001/chrome/chrome_syzygy.gyp
File chrome/chrome_syzygy.gyp (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/chrome_syzygy.gyp#new...
chrome/chrome_syzygy.gyp:6: ['OS=="win" & and fastbuild==0', {
On 2011/11/28 15:09:27, Roger McFarlane (Chromium) wrote:
> & ?

Yeah :(. Fixed.

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/creat...
File chrome/tools/build/win/create_installer_archive.py (right):

http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/creat...
chrome/tools/build/win/create_installer_archive.py:401: print sys.argv
On 2011/11/28 15:09:27, Roger McFarlane (Chromium) wrote:
> a remnant debugging aid?

No, this was present in the original script. I suppose it helps with debugging
on the bots.

[robertshield, 2011-11-28 20:16:29]

On 2011/11/28 18:55:36, Ruðrugis wrote:
> Flushing dangling comments...
> 
> http://codereview.chromium.org/8698009/diff/2001/chrome/chrome_syzygy.gyp
> File chrome/chrome_syzygy.gyp (right):
> 
>
http://codereview.chromium.org/8698009/diff/2001/chrome/chrome_syzygy.gyp#new...
> chrome/chrome_syzygy.gyp:6: ['OS=="win" & and fastbuild==0', {
> On 2011/11/28 15:09:27, Roger McFarlane (Chromium) wrote:
> > & ?
> 
> Yeah :(. Fixed.
> 
>
http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/creat...
> File chrome/tools/build/win/create_installer_archive.py (right):
> 
>
http://codereview.chromium.org/8698009/diff/2001/chrome/tools/build/win/creat...
> chrome/tools/build/win/create_installer_archive.py:401: print sys.argv
> On 2011/11/28 15:09:27, Roger McFarlane (Chromium) wrote:
> > a remnant debugging aid?
> 
> No, this was present in the original script. I suppose it helps with debugging
> on the bots.

Yes, it's extremely helpful to have the build logs include the exact command
line each script was run with. Makes repro'ing build failures 743x easier.

[Nico, 2014-04-03 18:34:48]

https://codereview.chromium.org/8698009/diff/8001/chrome/installer/mini_insta...
File chrome/installer/mini_installer_syzygy.gyp (right):

https://codereview.chromium.org/8698009/diff/8001/chrome/installer/mini_insta...
chrome/installer/mini_installer_syzygy.gyp:8: 'lastchange_path':
'<(SHARED_INTERMEDIATE_DIR)/build/LASTCHANGE',
Are the variables up here used for anything?