Add psutil build step to fix pyauto media issues. Upgrade psutil to 0.4.0.

Fixes broken media tests which were dependent on psutil. I've only setup
the build rules for linux right now as we've only got a Linux bot.

The psutil build requires that it be run from source directory, so a
small wrapper script was necessary to handle this and convert output
paths to absolute. If there's a way to do this in GYP let me know.

Most of the CL is just psutil upgrades, the important files are:
chrome/chrome_tests.gypi
third_party/psutil/build.py

I attached the new build step to pyautolib. Let me know if it should be 
elsewhere.

BUG=106105
TEST=Ran linux build. Ran media pyauto test.

[DaleCurtis, 2011-12-01 22:17:42]

PTAL. Please suggest alternate reviewers if I've chosen poorly.

[Nirnimesh, 2011-12-07 00:30:41]

http://codereview.chromium.org/8774018/diff/1/chrome/chrome_tests.gypi
File chrome/chrome_tests.gypi (right):

http://codereview.chromium.org/8774018/diff/1/chrome/chrome_tests.gypi#newcod...
chrome/chrome_tests.gypi:4065: 'test_support_psutil',
where did you specify this dep on linux only?

FWIW, I've checked that psutil builds and runs fine on mac and win as well. I
won't mind enabling it for all, even though you only have a linux bot at
present.

http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py
File third_party/psutil/build.py (right):

http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py#new...
third_party/psutil/build.py:15: print 'usage: build.py <output dir>'
print >>sys.stderr

s/usage/Usage/

http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py#new...
third_party/psutil/build.py:22: ['python', 'setup.py', 'build', '--build-lib',
abs_out_path],
--build-lib still produces 'build' dir in cwd. You should not create any files
in src tree, as it confuses svn.

So, also point --build-temp to a temp path.

[DaleCurtis, 2011-12-07 01:26:08]

http://codereview.chromium.org/8774018/diff/1/chrome/chrome_tests.gypi
File chrome/chrome_tests.gypi (right):

http://codereview.chromium.org/8774018/diff/1/chrome/chrome_tests.gypi#newcod...
chrome/chrome_tests.gypi:4065: 'test_support_psutil',
On 2011/12/07 00:30:41, Nirnimesh wrote:
> where did you specify this dep on linux only?
> 
> FWIW, I've checked that psutil builds and runs fine on mac and win as well. I
> won't mind enabling it for all, even though you only have a linux bot at
> present.

In the conditions section above, I only provided outputs for OS=="linux". If you
tell me what outputs were produced by the windows/mac builds I'll add sections
for them.

http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py
File third_party/psutil/build.py (right):

http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py#new...
third_party/psutil/build.py:15: print 'usage: build.py <output dir>'
On 2011/12/07 00:30:41, Nirnimesh wrote:
> print >>sys.stderr
> 
> s/usage/Usage/

Done.

http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py#new...
third_party/psutil/build.py:22: ['python', 'setup.py', 'build', '--build-lib',
abs_out_path],
On 2011/12/07 00:30:41, Nirnimesh wrote:
> --build-lib still produces 'build' dir in cwd. You should not create any files
> in src tree, as it confuses svn.
> 
> So, also point --build-temp to a temp path.

Done.

[Nirnimesh, 2011-12-07 02:56:05]

>
>
> In the conditions section above, I only provided outputs for
> OS=="linux". If you tell me what outputs were produced by the
> windows/mac builds I'll add sections for them.


When using build --build-lib, on Mac I have:

/tmp/bb/_psutil_osx.so
/tmp/bb/_psutil_posix.so
/tmp/bb/psutil
/tmp/bb/psutil/__init__.py
/tmp/bb/psutil/_common.py
/tmp/bb/psutil/_compat.py
/tmp/bb/psutil/_psbsd.py
/tmp/bb/psutil/_pslinux.py
/tmp/bb/psutil/_psmswindows.py
/tmp/bb/psutil/_psosx.py
/tmp/bb/psutil/_psposix.py
/tmp/bb/psutil/error.py
/tmp/bb/psutil/setup.py

Win has _psutil_mswindows.pyd instead of the .so's
The .py files are the same.


>
>
> http://codereview.chromium.**org/8774018/diff/1/third_**
>
party/psutil/build.py<http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py>
> File third_party/psutil/build.py (right):
>
> http://codereview.chromium.**org/8774018/diff/1/third_**
>
party/psutil/build.py#**newcode15<http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py#newcode15>
> third_party/psutil/build.py:**15: print 'usage: build.py <output dir>'
> On 2011/12/07 00:30:41, Nirnimesh wrote:
>
>> print >>sys.stderr
>>
>
>  s/usage/Usage/
>>
>
> Done.
>
>
> http://codereview.chromium.**org/8774018/diff/1/third_**
>
party/psutil/build.py#**newcode22<http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py#newcode22>
> third_party/psutil/build.py:**22: ['python', 'setup.py', 'build',
> '--build-lib', abs_out_path],
> On 2011/12/07 00:30:41, Nirnimesh wrote:
>
>> --build-lib still produces 'build' dir in cwd. You should not create
>>
> any files
>
>> in src tree, as it confuses svn.
>>
>
>  So, also point --build-temp to a temp path.
>>
>
> Done.
>
>
http://codereview.chromium.**org/8774018/<http://codereview.chromium.org/8774...
>



-- 
../NiR

[DaleCurtis, 2011-12-07 23:22:48]

When you run build.py on Linux do you get a setup.py as well? Seems weird that
is only in the mac/win variants.

Otherwise I've added the mac/win handlers.

On 2011/12/07 02:56:05, Nirnimesh wrote:
> >
> >
> > In the conditions section above, I only provided outputs for
> > OS=="linux". If you tell me what outputs were produced by the
> > windows/mac builds I'll add sections for them.
> 
> 
> When using build --build-lib, on Mac I have:
> 
> /tmp/bb/_psutil_osx.so
> /tmp/bb/_psutil_posix.so
> /tmp/bb/psutil
> /tmp/bb/psutil/__init__.py
> /tmp/bb/psutil/_common.py
> /tmp/bb/psutil/_compat.py
> /tmp/bb/psutil/_psbsd.py
> /tmp/bb/psutil/_pslinux.py
> /tmp/bb/psutil/_psmswindows.py
> /tmp/bb/psutil/_psosx.py
> /tmp/bb/psutil/_psposix.py
> /tmp/bb/psutil/error.py
> /tmp/bb/psutil/setup.py
> 
> Win has _psutil_mswindows.pyd instead of the .so's
> The .py files are the same.
> 
> 
> >
> >
> > http://codereview.chromium.**org/8774018/diff/1/third_**
> >
>
party/psutil/build.py<http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py>
> > File third_party/psutil/build.py (right):
> >
> > http://codereview.chromium.**org/8774018/diff/1/third_**
> >
>
party/psutil/build.py#**newcode15<http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py#newcode15>
> > third_party/psutil/build.py:**15: print 'usage: build.py <output dir>'
> > On 2011/12/07 00:30:41, Nirnimesh wrote:
> >
> >> print >>sys.stderr
> >>
> >
> >  s/usage/Usage/
> >>
> >
> > Done.
> >
> >
> > http://codereview.chromium.**org/8774018/diff/1/third_**
> >
>
party/psutil/build.py#**newcode22<http://codereview.chromium.org/8774018/diff/1/third_party/psutil/build.py#newcode22>
> > third_party/psutil/build.py:**22: ['python', 'setup.py', 'build',
> > '--build-lib', abs_out_path],
> > On 2011/12/07 00:30:41, Nirnimesh wrote:
> >
> >> --build-lib still produces 'build' dir in cwd. You should not create
> >>
> > any files
> >
> >> in src tree, as it confuses svn.
> >>
> >
> >  So, also point --build-temp to a temp path.
> >>
> >
> > Done.
> >
> >
>
http://codereview.chromium.**org/8774018/%3Chttp://codereview.chromium.org/87...>
> >
> 
> 
> 
> -- 
> ../NiR

[Nirnimesh, 2011-12-08 00:23:41]

Yes, I do get setup.py on mac and win.
But you don't need to care about it, since it's unnecessary.

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi
File chrome/chrome_tests.gypi (right):

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
chrome/chrome_tests.gypi:446: {
I forgot to point out earlier: the gyp rule for psutil should be in third_party,
in its own file.
Then you can depend on it with: ../third_party/psutil/psutil.gyp:psutil

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
chrome/chrome_tests.gypi:464: '../third_party/psutil/psutil/_psutil_posix.h',
please include setup.py as well, since we want to re-build if it changes.

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
chrome/chrome_tests.gypi:484: },],
nit: first , not necessary

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
chrome/chrome_tests.gypi:501: 'message': 'Generating psutil binary',
s/binary/binaries/

[DaleCurtis, 2011-12-08 22:04:38]

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi
File chrome/chrome_tests.gypi (right):

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
chrome/chrome_tests.gypi:446: {
On 2011/12/08 00:23:41, Nirnimesh wrote:
> I forgot to point out earlier: the gyp rule for psutil should be in
third_party,
> in its own file.
> Then you can depend on it with: ../third_party/psutil/psutil.gyp:psutil

Hurm. So I'm having trouble getting this working without warnings and errors
being thrown.

I had to set 'type': 'none' and I'm still getting a "ninja: WARNING: multiple
rules generate psutil. build will not be correct; continuing anyway" error. Any
ideas?

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
chrome/chrome_tests.gypi:464: '../third_party/psutil/psutil/_psutil_posix.h',
On 2011/12/08 00:23:41, Nirnimesh wrote:
> please include setup.py as well, since we want to re-build if it changes.

Done.

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
chrome/chrome_tests.gypi:484: },],
On 2011/12/08 00:23:41, Nirnimesh wrote:
> nit: first , not necessary

Done.

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
chrome/chrome_tests.gypi:501: 'message': 'Generating psutil binary',
On 2011/12/08 00:23:41, Nirnimesh wrote:
> s/binary/binaries/

Done.

[cmp, 2011-12-08 22:30:40]

I trust Nirnimesh has this covered.  Let me know if you want me to take a look
at something here.

[Nirnimesh, 2011-12-08 22:40:55]

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi
File chrome/chrome_tests.gypi (right):

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
chrome/chrome_tests.gypi:446: {
On 2011/12/08 22:04:38, DaleCurtis wrote:
> On 2011/12/08 00:23:41, Nirnimesh wrote:
> > I forgot to point out earlier: the gyp rule for psutil should be in
> third_party,
> > in its own file.
> > Then you can depend on it with: ../third_party/psutil/psutil.gyp:psutil
> 
> Hurm. So I'm having trouble getting this working without warnings and errors
> being thrown.
> 
> I had to set 'type': 'none' and I'm still getting a "ninja: WARNING: multiple
> rules generate psutil. build will not be correct; continuing anyway" error.
Any
> ideas?

Can you upload your latest CL?
You could take other packages in third_party as examples. say libjpeg

[DaleCurtis, 2011-12-08 23:03:36]

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi
File chrome/chrome_tests.gypi (right):

http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
chrome/chrome_tests.gypi:446: {
On 2011/12/08 22:40:56, Nirnimesh wrote:
> On 2011/12/08 22:04:38, DaleCurtis wrote:
> > On 2011/12/08 00:23:41, Nirnimesh wrote:
> > > I forgot to point out earlier: the gyp rule for psutil should be in
> > third_party,
> > > in its own file.
> > > Then you can depend on it with: ../third_party/psutil/psutil.gyp:psutil
> > 
> > Hurm. So I'm having trouble getting this working without warnings and errors
> > being thrown.
> > 
> > I had to set 'type': 'none' and I'm still getting a "ninja: WARNING:
multiple
> > rules generate psutil. build will not be correct; continuing anyway" error.
> Any
> > ideas?
> 
> Can you upload your latest CL?
> You could take other packages in third_party as examples. say libjpeg

Done. I looked there, but psutil is a unique in that it generates objects that
aren't supposed to be linked with anything. With a type other than none gyp
tries to link the .so's against pyautolib.

[Nirnimesh, 2011-12-09 00:00:40]

On 2011/12/08 23:03:36, DaleCurtis wrote:
> http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi
> File chrome/chrome_tests.gypi (right):
> 
>
http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
> chrome/chrome_tests.gypi:446: {
> On 2011/12/08 22:40:56, Nirnimesh wrote:
> > On 2011/12/08 22:04:38, DaleCurtis wrote:
> > > On 2011/12/08 00:23:41, Nirnimesh wrote:
> > > > I forgot to point out earlier: the gyp rule for psutil should be in
> > > third_party,
> > > > in its own file.
> > > > Then you can depend on it with: ../third_party/psutil/psutil.gyp:psutil
> > > 
> > > Hurm. So I'm having trouble getting this working without warnings and
errors
> > > being thrown.
> > > 
> > > I had to set 'type': 'none' and I'm still getting a "ninja: WARNING:
> multiple
> > > rules generate psutil. build will not be correct; continuing anyway"
error.
> > Any
> > > ideas?
> > 
> > Can you upload your latest CL?
> > You could take other packages in third_party as examples. say libjpeg
> 
> Done. I looked there, but psutil is a unique in that it generates objects that
> aren't supposed to be linked with anything. With a type other than none gyp
> tries to link the .so's against pyautolib.

There are several examples of 'type': 'none' in chrome/chrome.gyp

I'm not sure what else is missing. You'd need to run it by a gyp expert.
mmentovai/bradnelson

[DaleCurtis, 2011-12-09 01:25:33]

On 2011/12/09 00:00:40, Nirnimesh wrote:
> On 2011/12/08 23:03:36, DaleCurtis wrote:
> > http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi
> > File chrome/chrome_tests.gypi (right):
> > 
> >
>
http://codereview.chromium.org/8774018/diff/10001/chrome/chrome_tests.gypi#ne...
> > chrome/chrome_tests.gypi:446: {
> > On 2011/12/08 22:40:56, Nirnimesh wrote:
> > > On 2011/12/08 22:04:38, DaleCurtis wrote:
> > > > On 2011/12/08 00:23:41, Nirnimesh wrote:
> > > > > I forgot to point out earlier: the gyp rule for psutil should be in
> > > > third_party,
> > > > > in its own file.
> > > > > Then you can depend on it with:
../third_party/psutil/psutil.gyp:psutil
> > > > 
> > > > Hurm. So I'm having trouble getting this working without warnings and
> errors
> > > > being thrown.
> > > > 
> > > > I had to set 'type': 'none' and I'm still getting a "ninja: WARNING:
> > multiple
> > > > rules generate psutil. build will not be correct; continuing anyway"
> error.
> > > Any
> > > > ideas?
> > > 
> > > Can you upload your latest CL?
> > > You could take other packages in third_party as examples. say libjpeg
> > 
> > Done. I looked there, but psutil is a unique in that it generates objects
that
> > aren't supposed to be linked with anything. With a type other than none gyp
> > tries to link the .so's against pyautolib.
> 
> There are several examples of 'type': 'none' in chrome/chrome.gyp
> 
> I'm not sure what else is missing. You'd need to run it by a gyp expert.
> mmentovai/bradnelson

Okay, I fixed the warnings. It didn't like the target being named psutil. Ninja
ended up generating a name conflict some how. I changed the name to psutil_lib
and everything seems okay.

In testing I noticed that pyautolib would now rebuild every time, even if
nothing had changed. This is because it was copying files over with time stamps
as is, and GYP got confused about whether it had generated psutil or not.

I updated the wrapper to touch all the Python related files now and GYP seems
happy.

PTAL.

[Nirnimesh, 2011-12-09 02:31:09]

LGTM

[commit-bot: I haz the power, 2011-12-09 19:29:42]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dalecurtis@chromium.org/8774018/19001

[commit-bot: I haz the power, 2011-12-09 20:34:32]

Try job failure for 8774018-19001 (retry) on mac_rel for step "compile" (clobber
build).
It's a second try, previously, step "compile" failed.
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=mac_rel&nu...

[DaleCurtis, 2011-12-09 21:22:14]

On 2011/12/09 20:34:32, I haz the power (commit-bot) wrote:
> Try job failure for 8774018-19001 (retry) on mac_rel for step "compile"
(clobber
> build).
> It's a second try, previously, step "compile" failed.
>
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=mac_rel&nu...

Not sure what's going on here, the logs look like goma blew up.

[commit-bot: I haz the power, 2011-12-09 21:53:07]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dalecurtis@chromium.org/8774018/19001

[commit-bot: I haz the power, 2011-12-09 23:13:08]

Try job failure for 8774018-19001 (retry) on mac_rel for step "compile" (clobber
build).
It's a second try, previously, step "compile" failed.
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=mac_rel&nu...

[DaleCurtis, 2011-12-09 23:14:16]

On 2011/12/09 23:13:08, I haz the power (commit-bot) wrote:
> Try job failure for 8774018-19001 (retry) on mac_rel for step "compile"
(clobber
> build).
> It's a second try, previously, step "compile" failed.
>
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=mac_rel&nu...

Sent email to goma-users, I'm not sure what's going on here. If I don' t hear
back by next week, I'll drop the mac build step.

[Nico, 2011-12-12 18:20:09]

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/build.py
File third_party/psutil/build.py (right):

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/build.py...
third_party/psutil/build.py:20: # the source directory, we must convert the
output path to an absolute path.
Can you use a _files prefix to let gyp do this transform for you?

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.gyp
File third_party/psutil/psutil.gyp (right):

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.g...
third_party/psutil/psutil.gyp:83: 'build.py',
Shelling out to a script to do a compilation is…unusual. Have you considered
doing a real gyp file, so that the build can be done by a build system instead
of by a python script?

[DaleCurtis, 2011-12-12 19:23:20]

PTAL, I've disable the Mac builds for now.

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/build.py
File third_party/psutil/build.py (right):

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/build.py...
third_party/psutil/build.py:20: # the source directory, we must convert the
output path to an absolute path.
On 2011/12/12 18:20:12, Nico wrote:
> Can you use a _files prefix to let gyp do this transform for you?

If so, I couldn't get it to work. The problem is we need <(PRODUCT_DIR) to be an
absolute path. Dumping that into a var with any of the special prefixes here
didn't help:

http://code.google.com/p/gyp/wiki/InputFormatReference#Pathname_Relativization

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.gyp
File third_party/psutil/psutil.gyp (right):

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.g...
third_party/psutil/psutil.gyp:83: 'build.py',
Hi Nico, I'm not sure what you mean by a real gyp file. Can you elaborate? Do
you mean converting the setup.py that's packaged with psutil into a gyp file? If
so, I'd rather not go that way since it's subject to upstream changes.

On 2011/12/12 18:20:12, Nico wrote:
> Shelling out to a script to do a compilation is…unusual. Have you considered
> doing a real gyp file, so that the build can be done by a build system instead
> of by a python script?

[Nirnimesh, 2011-12-12 21:07:10]

Please update the description.

http://codereview.chromium.org/8774018/diff/26001/third_party/psutil/build.py
File third_party/psutil/build.py (right):

http://codereview.chromium.org/8774018/diff/26001/third_party/psutil/build.py...
third_party/psutil/build.py:13: 
nit: leave another blank line here

http://codereview.chromium.org/8774018/diff/26001/third_party/psutil/build.py...
third_party/psutil/build.py:16: print >> sys.stderr, 'Usage: build.py <output
dir>'
nit: remove space after >>

[Nico, 2011-12-12 23:20:43]

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.gyp
File third_party/psutil/psutil.gyp (right):

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.g...
third_party/psutil/psutil.gyp:83: 'build.py',
On 2011/12/12 19:23:20, DaleCurtis wrote:
> Hi Nico, I'm not sure what you mean by a real gyp file. Can you elaborate? Do
> you mean converting the setup.py that's packaged with psutil into a gyp file?

Yes.

> If so, I'd rather not go that way since it's subject to upstream changes.

It's what we do for all third-party libraries we pull. Usually, upstream doesn't
change that often, and keeping source lists in sync isn't that big of a deal.

Having a normal gyp file means that stuff like goma, custom compilers, new build
systems, etc all have a chance of being supported.

I don't know if it's possible to do python packages that way though. Ideally,
the media tests would not depend on psutil – do you know what they use it for?

[DaleCurtis, 2011-12-12 23:34:56]

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.gyp
File third_party/psutil/psutil.gyp (right):

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.g...
third_party/psutil/psutil.gyp:83: 'build.py',
I didn't realize you guys were already doing that, if that's the case I can look
into it. At least for the Linux build (which is all we need right now) it looks
pretty easy.

The tests are using psutil to record process utilization metrics while videos
are playing. We could remove it from the tree and just require users install the
package prior to use.

That would mean the chromiumos.perf_av build slave would need these packages
installed manually.

On 2011/12/12 23:20:43, Nico wrote:
> On 2011/12/12 19:23:20, DaleCurtis wrote:
> > Hi Nico, I'm not sure what you mean by a real gyp file. Can you elaborate?
Do
> > you mean converting the setup.py that's packaged with psutil into a gyp
file?
> 
> Yes.
> 
> > If so, I'd rather not go that way since it's subject to upstream changes.
> 
> It's what we do for all third-party libraries we pull. Usually, upstream
doesn't
> change that often, and keeping source lists in sync isn't that big of a deal.
> 
> Having a normal gyp file means that stuff like goma, custom compilers, new
build
> systems, etc all have a chance of being supported.
> 
> I don't know if it's possible to do python packages that way though. Ideally,
> the media tests would not depend on psutil – do you know what they use it for?

[Nico, 2011-12-12 23:42:11]

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.gyp
File third_party/psutil/psutil.gyp (right):

http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.g...
third_party/psutil/psutil.gyp:83: 'build.py',
On 2011/12/12 23:34:57, DaleCurtis wrote:
> I didn't realize you guys were already doing that, if that's the case I can
look
> into it. At least for the Linux build (which is all we need right now) it
looks
> pretty easy.
> 
> The tests are using psutil to record process utilization metrics while videos
> are playing. We could remove it from the tree and just require users install
the
> package prior to use.

It sounds like those are all chrome processes – could you use chrome's task
manager code for that instead?

> That would mean the chromiumos.perf_av build slave would need these packages
> installed manually.

If this is just for linux, you could add the package to
build/install-build-deps.sh, which is executed on new slaves. (I think
chromiumos uses it, too.) chrome-infra-team can install packages on existing
bots.

That sounds a tad cleaner to me (this! is! not! gentoo! :-) ), but maybe you
should ask chromium-dev what the right thing to do is here. (If the chrome task
manager code can do what you want, that's probably the way to go.)

[DaleCurtis, 2011-12-12 23:48:15]

On 2011/12/12 23:34:56, DaleCurtis wrote:
>
http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.gyp
> File third_party/psutil/psutil.gyp (right):
> 
>
http://codereview.chromium.org/8774018/diff/19001/third_party/psutil/psutil.g...
> third_party/psutil/psutil.gyp:83: 'build.py',
> I didn't realize you guys were already doing that, if that's the case I can
look
> into it. At least for the Linux build (which is all we need right now) it
looks
> pretty easy.
> 
> The tests are using psutil to record process utilization metrics while videos
> are playing. We could remove it from the tree and just require users install
the
> package prior to use.
> 
> That would mean the chromiumos.perf_av build slave would need these packages
> installed manually.
> 
> On 2011/12/12 23:20:43, Nico wrote:
> > On 2011/12/12 19:23:20, DaleCurtis wrote:
> > > Hi Nico, I'm not sure what you mean by a real gyp file. Can you elaborate?
> Do
> > > you mean converting the setup.py that's packaged with psutil into a gyp
> file?
> > 
> > Yes.
> > 
> > > If so, I'd rather not go that way since it's subject to upstream changes.
> > 
> > It's what we do for all third-party libraries we pull. Usually, upstream
> doesn't
> > change that often, and keeping source lists in sync isn't that big of a
deal.
> > 
> > Having a normal gyp file means that stuff like goma, custom compilers, new
> build
> > systems, etc all have a chance of being supported.
> > 
> > I don't know if it's possible to do python packages that way though.
Ideally,
> > the media tests would not depend on psutil – do you know what they use it
for?

Just talked to cmp and he said install-build-deps is fine. I'm going to abandon
this CL, submit a new one to remove psutil from the tree and update
install-build-deps.

Thanks for the reviews everyone!