Terminate the engine if another instance is already running

Contrary to what we discussed, I didn't use named mutexes for this. I'm using WaitNamedPipe to check for the existence of a pipe with that name. Works fine, and is simpler.

[Wladimir Palant, 2013-06-07 06:16:08]

http://codereview.adblockplus.org/10809053/diff/1/src/engine/main.cpp
File src/engine/main.cpp (right):

http://codereview.adblockplus.org/10809053/diff/1/src/engine/main.cpp#newcode194
src/engine/main.cpp:194: }
That's causing a race condition. It can take several seconds for this engine
instance to create its own pipe - in the meantime additional engine instances
are free to start.

If it doesn't look like we can avoid race conditions with pipes then we should
definitely use mutexes.

[Felix Dahlke, 2013-06-07 06:37:22]

http://codereview.adblockplus.org/10809053/diff/1/src/engine/main.cpp
File src/engine/main.cpp (right):

http://codereview.adblockplus.org/10809053/diff/1/src/engine/main.cpp#newcode194
src/engine/main.cpp:194: }
On 2013/06/07 06:16:08, Wladimir Palant wrote:
> That's causing a race condition. It can take several seconds for this engine
> instance to create its own pipe - in the meantime additional engine instances
> are free to start.
> 
> If it doesn't look like we can avoid race conditions with pipes then we should
> definitely use mutexes.

Hm, you're right, this actually shouldn't work. Makes me wonder why it does... I
spawned about 5 engine instances in parallel several times, four of those always
exited after a few ms. Never saw two engines running after this was added.

However, I guess named mutexes are better after all. Even if we create the pipe
right away and the filterEngine later, we could in theory have a race condition
here.

[Felix Dahlke, 2013-06-07 08:07:55]

Implemented this with named pipes. It's in a new review, since this change was
already pushed:

http://codereview.adblockplus.org/10891004/

I'll close this one.