[Webkit-unassigned] [Bug 61287] [UNIX] Don't use SOCK_DGRAM in socketpair()

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=61287





--- Comment #20 from Carlos Garcia Campos <cgarcia at igalia.com>  2011-05-26 03:23:44 PST ---
(In reply to comment #19)
> (In reply to comment #17)
> > And the same code would work on windows too.
> 
> I don't think it's the correct solution to windows IPC.

Yes, you are right, and I agree using low level apis might be more robust even if it's more code and I think it's good sharing code. 

When you say the message is buffered redundantly, what do you mean? Comparing current code and the patch we have the followingf heap allocations:

 - Current code: We allocate the readBuffer in platformInitialize(), then everytime readyReadHandler() is called
  + attachmentDescriptorBuffer is allocated/deallocated
  + attachmentSizes is allocated/deallocated
  + fileDescriptors is allocated/deallocated

 - With the patch: We allocate readBuffer and fileDescriptors buffers in platformInitialize(), then everytime readyReadHandler() is called:
  + attachmentDescriptorBuffer is allocated/deallocated in readBytesFromSocket()
  + attachmentSizes  is allocated/deallocated in processMessage()

So we are even reducing the number of allocations/deallocations. We could avoid also attachmentDescriptorBuffer allocation in both cases using alloca() if available.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.