[Bug 17195] NamedPipe datagrams need to be _really_ datagrams

[wine-bugs at winehq.org]

http://bugs.winehq.org/show_bug.cgi?id=17195

Adam Martinson <adam.r.martinson at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #38210|0                           |1
        is obsolete|                            |

--- Comment #64 from Adam Martinson <adam.r.martinson at gmail.com> 2012-01-07 19:05:16 CST ---
Created attachment 38295
  --> http://bugs.winehq.org/attachment.cgi?id=38295
patch (work in progress)

Ok, so Alexandre says we don't want to use shm; he suggested sending the
messages through the server as part of the request/reply data, and... it works!
 It passes all of the python multiprocessing tests anyhow (tested 2.7 and 3.2).

However, there's a race condition with RTL_RWLOCKs somewhere, I'm not sure if
it's related to my changes or not.  In the python tests it manifests as an
occasional hang at

test_lock (__main__.WithManagerTestLock) ...

In general I get

err:ntdll:RtlDeleteResource Deleting active MRSW lock (0x112414), expect
failure

as the last message occasionally regardless of what wine is running, I'm
guessing the 2 are related. Any insight into that would be helpful.

-- 
Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email
Do not reply to this email, post in Bugzilla using the
above URL to reply.
------- You are receiving this mail because: -------
You are watching all bug changes.