Get service processes working on Mac and Linux.

Get service processes working on Mac and Linux.

Service processes will now launch, and attach to ipc channels.

BUG=NONE
TEST=BUILD

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=73425

[dmac, 2011-02-01 22:42:52]

This should get service processes running on Mac and Linux.

I have beefed up the unittests. Autorun is not done, and will need to be sorted
out. Also we are currently launching the service as a child process, and will
most likely want to change this on Mac and Linux.

Sorry for how long this took. Turned out to be rather painful, although the
final implementation is pretty simple.

[Scott Byer, 2011-02-02 00:12:37]

http://codereview.chromium.org/6349029/diff/4001/chrome/common/service_proces...
File chrome/common/service_process_util_unittest.cc (right):

http://codereview.chromium.org/6349029/diff/4001/chrome/common/service_proces...
chrome/common/service_process_util_unittest.cc:42: MessageLoop* IOMessageLoop()
{ return io_thread_.message_loop(); };
nit: lint error

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
File chrome/common/service_process_util.cc (right):

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
chrome/common/service_process_util.cc:129: scoped_ptr<base::SharedMemory>
shared_mem_service_data;
Did this function need to get moved? It's not like the functions are in header
declaration order before or after the move, and it would help minimize
differences back in the old location.

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
File chrome/common/service_process_util_posix.cc (right):

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
chrome/common/service_process_util_posix.cc:32: // Attempts to take a lock named
|name|. If |waiting| then is will
nit: then this will

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
chrome/common/service_process_util_posix.cc:96: if ((length ==
sizeof(kShutDownMessage)) && (buffer == kShutDownMessage)) {
Should be sizeof(buffer) to match the write side.

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
chrome/common/service_process_util_posix.cc:241: return running_lock.get() ==
NULL;
Shouldn't this just call CheckServiceProcessReady()?

[dmac, 2011-02-02 00:56:24]

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
File chrome/common/service_process_util.cc (right):

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
chrome/common/service_process_util.cc:129: scoped_ptr<base::SharedMemory>
shared_mem_service_data;
On 2011/02/02 00:12:38, Scott Byer wrote:
> Did this function need to get moved? It's not like the functions are in header
> declaration order before or after the move, and it would help minimize
> differences back in the old location.

Yes, I needed to take it out of the anonymous namespace that it was in.

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
File chrome/common/service_process_util_posix.cc (right):

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
chrome/common/service_process_util_posix.cc:32: // Attempts to take a lock named
|name|. If |waiting| then is will
On 2011/02/02 00:12:38, Scott Byer wrote:
> nit: then this will

Done.

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
chrome/common/service_process_util_posix.cc:96: if ((length ==
sizeof(kShutDownMessage)) && (buffer == kShutDownMessage)) {
On 2011/02/02 00:12:38, Scott Byer wrote:
> Should be sizeof(buffer) to match the write side.

Done.

http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
chrome/common/service_process_util_posix.cc:241: return running_lock.get() ==
NULL;
On 2011/02/02 00:12:38, Scott Byer wrote:
> Shouldn't this just call CheckServiceProcessReady()?

Got rid of it instead.

[Scott Byer, 2011-02-02 01:12:19]

LGTM.

On 2011/02/02 00:56:24, dmac wrote:
>
http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
> File chrome/common/service_process_util.cc (right):
> 
>
http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
> chrome/common/service_process_util.cc:129: scoped_ptr<base::SharedMemory>
> shared_mem_service_data;
> On 2011/02/02 00:12:38, Scott Byer wrote:
> > Did this function need to get moved? It's not like the functions are in
header
> > declaration order before or after the move, and it would help minimize
> > differences back in the old location.
> 
> Yes, I needed to take it out of the anonymous namespace that it was in.
> 
>
http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
> File chrome/common/service_process_util_posix.cc (right):
> 
>
http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
> chrome/common/service_process_util_posix.cc:32: // Attempts to take a lock
named
> |name|. If |waiting| then is will
> On 2011/02/02 00:12:38, Scott Byer wrote:
> > nit: then this will
> 
> Done.
> 
>
http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
> chrome/common/service_process_util_posix.cc:96: if ((length ==
> sizeof(kShutDownMessage)) && (buffer == kShutDownMessage)) {
> On 2011/02/02 00:12:38, Scott Byer wrote:
> > Should be sizeof(buffer) to match the write side.
> 
> Done.
> 
>
http://codereview.chromium.org/6349029/diff/2005/chrome/common/service_proces...
> chrome/common/service_process_util_posix.cc:241: return running_lock.get() ==
> NULL;
> On 2011/02/02 00:12:38, Scott Byer wrote:
> > Shouldn't this just call CheckServiceProcessReady()?
> 
> Got rid of it instead.