Changes necessary to compile gpu demos on linux. It is not functional yet jus...

Changes necessary to compile gpu demos on linux. It is not functional yet just compiling. 
BUG=26099
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=37977

[alokp, 2010-01-29 22:46:32]

I need help with -fPIC flag. My pepper demos depend on base and gpu that are not
compiled with -fPIC and compiler is complaining about it.

http://codereview.chromium.org/552240/diff/1/4
File gpu/demos/demos.gyp (right):

http://codereview.chromium.org/552240/diff/1/4#newcode82
gpu/demos/demos.gyp:82: 'cflags': ['-fPIC'],
-fPIC is still not working. It is complaining about a few files in base and gpu
needing dynamic reloc support. Do we need to compile all libraries with -fPIC?

[apatrick_chromium, 2010-01-29 22:57:21]

lgtm for windows

http://codereview.chromium.org/552240/diff/1/5
File gpu/demos/framework/window.cc (right):

http://codereview.chromium.org/552240/diff/1/5#newcode38
gpu/demos/framework/window.cc:38: if (window_handle_ == NULL) return false;
can you put the returns on the next line so it is easier to set breakpoints?

[piman, 2010-01-30 00:24:19]

http://codereview.chromium.org/552240/diff/1/2
File gpu/command_buffer/common/cmd_buffer_common.cc (right):

http://codereview.chromium.org/552240/diff/1/2#newcode12
gpu/command_buffer/common/cmd_buffer_common.cc:12: // even though it is a static
const int variable.
gcc is fine if you define it in the header, but you still need to declare
storage in the .cc (without initialization).
That's actually what the C++ spec wants.

However MSVC freaks out if you do respect the spec and declare the storage, so
you usually end up #ifdef-ing it.

http://codereview.chromium.org/552240/diff/1/4
File gpu/demos/demos.gyp (right):

http://codereview.chromium.org/552240/diff/1/4#newcode82
gpu/demos/demos.gyp:82: 'cflags': ['-fPIC'],
On 2010/01/29 22:46:32, alokp wrote:
> -fPIC is still not working. It is complaining about a few files in base and
gpu
> needing dynamic reloc support. Do we need to compile all libraries with -fPIC?

On x64, yes. I added a way in chrome/common.gypi to compile everything with
-fPIC for the pepper plugin, but it's not turned on by default in Chrome because
it slows things down. If you specify GYP_DEFINES="linux_fpic=1" then re-run gyp,
it will compile everything with -fpic.

My suggestion is to disable compiling of all of this on linux/x64 unless
linux_fpic is 1.

[alokp, 2010-02-01 19:09:19]

Need help with the runhooks error.

http://codereview.chromium.org/552240/diff/1/2
File gpu/command_buffer/common/cmd_buffer_common.cc (right):

http://codereview.chromium.org/552240/diff/1/2#newcode12
gpu/command_buffer/common/cmd_buffer_common.cc:12: // even though it is a static
const int variable.
AFAIK the spec recently removed this requirement to facilitate writing template
classes using static const int. Otherwise they needed to use enums. Anyways this
change does not hurt anything.

http://codereview.chromium.org/552240/diff/1/4
File gpu/demos/demos.gyp (right):

http://codereview.chromium.org/552240/diff/1/4#newcode82
gpu/demos/demos.gyp:82: 'cflags': ['-fPIC'],
I have removed -fPIC from this gyp file because commn.gypi has everything I
want. I only need to set "linux_fpic=1" to make it work. Added a
enable_pepper_demos variable to disable pepper demos on linux/x64 if linux_fpic
is not 1. However runhooks is complaining about linxu_fpic not defined.

NameError: name 'linux_fpic' is not defined while evaluating condition
'OS=="linux" and (target_arch=="x64" or target_arch=="arm") and linux_fpic!=1'
in src/gpu/demos/demos.gyp while loading dependencies of src/build/all.gyp while
trying to load src/build/all.gyp

http://codereview.chromium.org/552240/diff/1/5
File gpu/demos/framework/window.cc (right):

http://codereview.chromium.org/552240/diff/1/5#newcode38
gpu/demos/framework/window.cc:38: if (window_handle_ == NULL) return false;
On 2010/01/29 22:57:22, apatrick_chromium wrote:
> can you put the returns on the next line so it is easier to set breakpoints?

Done.

[apatrick_chromium, 2010-02-01 19:44:01]

http://codereview.chromium.org/552240/diff/5001/5004
File gpu/demos/demos.gyp (right):

http://codereview.chromium.org/552240/diff/5001/5004#newcode12
gpu/demos/demos.gyp:12: 'variables': {
I don't know if this is causing the fpic problem but, because this is a python
dictionary, you can't have to 'variables' keys. It looks like people usually put
the 'variables' key before the 'includes' key. So I would move everything you
have in this block into the 'variables' block above.

[alokp, 2010-02-01 20:02:13]

http://codereview.chromium.org/552240/diff/5001/5004
File gpu/demos/demos.gyp (right):

http://codereview.chromium.org/552240/diff/5001/5004#newcode12
gpu/demos/demos.gyp:12: 'variables': {
I originally had only one variable blocks as you suggest and was having the same
problem. Since it was complaining about linux_fpic variable, which is defined in
common.gypi, I moved it after including common.gypi. But that did not fix it.

[bradn, 2010-02-02 00:13:23]

FYI

http://codereview.chromium.org/552240/diff/8001/8002
File gpu/demos/demos.gyp (right):

http://codereview.chromium.org/552240/diff/8001/8002#newcode20
gpu/demos/demos.gyp:20: },
You may need to elevate this up a level by adding:
'enable_pepper_demos%': '<(enable_pepper_demos)',
here.

http://codereview.chromium.org/552240/diff/8001/8007
File gpu/demos/framework/window_linux.cc (right):

http://codereview.chromium.org/552240/diff/8001/8007#newcode7
gpu/demos/framework/window_linux.cc:7: namespace {
leftover?

http://codereview.chromium.org/552240/diff/8001/8005
File gpu/demos/framework/window_mac.mm (right):

http://codereview.chromium.org/552240/diff/8001/8005#newcode7
gpu/demos/framework/window_mac.mm:7: namespace {
leftover?

http://codereview.chromium.org/552240/diff/8001/8013
File third_party/gles2_book/Common/Include/esUtil.h (left):

http://codereview.chromium.org/552240/diff/8001/8013#oldcode114
third_party/gles2_book/Common/Include/esUtil.h:114: //
Since this is third_party, do we need to update the README?

[alokp, 2010-02-02 17:12:45]

http://codereview.chromium.org/552240/diff/8001/8002
File gpu/demos/demos.gyp (right):

http://codereview.chromium.org/552240/diff/8001/8002#newcode20
gpu/demos/demos.gyp:20: },
did you mean this? It does not work either.

NameError: name 'target_arch' is not defined while evaluating condition
'OS=="linux" and (target_arch=="x64" or target_arch=="arm") and linux_fpic!=1'
in /usr/local/google/home/alokp/chrome/src/gpu/demos/demos.gyp while loading
dependencies of /usr/local/google/home/alokp/chrome/src/build/all.gyp while
trying to load src/build/all.gyp

http://codereview.chromium.org/552240/diff/8001/8007
File gpu/demos/framework/window_linux.cc (right):

http://codereview.chromium.org/552240/diff/8001/8007#newcode7
gpu/demos/framework/window_linux.cc:7: namespace {
removed

http://codereview.chromium.org/552240/diff/8001/8005
File gpu/demos/framework/window_mac.mm (right):

http://codereview.chromium.org/552240/diff/8001/8005#newcode7
gpu/demos/framework/window_mac.mm:7: namespace {
removed

http://codereview.chromium.org/552240/diff/8001/8013
File third_party/gles2_book/Common/Include/esUtil.h (left):

http://codereview.chromium.org/552240/diff/8001/8013#oldcode114
third_party/gles2_book/Common/Include/esUtil.h:114: //
Updated README.chromium

[Michael Moss, 2010-02-02 19:28:58]

http://codereview.chromium.org/552240/diff/13003/14016
File gpu/demos/demos.gyp (right):

http://codereview.chromium.org/552240/diff/13003/14016#newcode8
gpu/demos/demos.gyp:8: 'variables': {
If you take out this 'variables' sub-dict (leaving the 'conditions' below), and
...

http://codereview.chromium.org/552240/diff/13003/14016#newcode22
gpu/demos/demos.gyp:22: 'enable_pepper_demos%': '<(enable_pepper_demos)',
... remove this redefinition of 'enable_pepper_demos', then add 'linux_fpic' to
the 'variables' sub-dict in common.gypi (where things like
'linux_chromium_breakpad' are initially defined), I think it will take care of
your undefined variables problem. It's horribly convoluted and unintuitive, but
such is gyp variable handling :(.

[alokp, 2010-02-02 20:20:51]

Michael: You rock! This version works as I wanted but I have no clue what is
going on.