[Bug 17195] NamedPipe datagrams need to be _really_ datagrams

[wine-bugs at winehq.org]

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

--- Comment #162 from Sebastian Lackner <sebastian at fds-team.de> ---
When SO_PEEK_OFF is not supported, then my implementation will fall back to
SOCK_STREAM sockets. This is also the behaviour for kernels < 3.4. Without
having SO_PEEK_OFF its not possible to receive parts of a message, so it will
not be used at all, to prevent regressions. In this case the patches will only
fix the SetNamedPipeHandleState issue, but not all the other details (and will
print fixmes). For FreeBSD we would need a similar functionality. If someone
experienced with the FreeBSD kernel has some time, feel free to join
#wine-compholio (on freenode) to discuss the details about possible
implementations. We need either a completely different solution (if available),
or will have to reimplement SO_PEEK_OFF there.

BTW, short status update: The first four relatively small Wine patches (which
should be definitely correct) are currently waiting to be reviewed. Besides
that I'm just updating some later patches in the series, some additional tests
revealed that overlapped stuff doesn't work correctly yet, with both Adams and
my patchset. Moreover we're preparing some kernel patches, we discovered some
weird behaviour for zero byte reads, which can be considered a kernel bug... ;)

-- 
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.