PATCH multi-channel & fd passing
obnox at samba.org
Thu Mar 19 10:39:54 MDT 2015
On 2015-03-19 at 16:54 +0100, Michael Adam wrote:
> On 2015-03-19 at 13:27 +0100, Michael Adam wrote:
> > On 2015-03-19 at 12:09 +0100, Volker Lendecke wrote:
> > > On Thu, Mar 19, 2015 at 10:16:54AM +0000, Noel Power wrote:
> > > > it doesn't fix anything but a personal annoyance and something that I
> > > > found confusing (e.g. the anonymous block) which was adjacent to code I
> > > > was touching to fix some other problem. This was just an alternative to
> > > > the orig patch I posted with the fix above, really not worth having a
> > > > discussion about, sorry for the noise
> > >
> > > Ok. I pushed the other patch, thanks for that fix!
> > >
> > > I wonder how we survive the relevant torture tests....
> > I have an idea: I think our tests only ever triggered
> > the complicated framentation/queing code path, but not
> > the fastpath that sends a small message directly, which
> > is the code path fixed with Noel's patch.
> > Currently working on extending our torture code to
> > add a regression test for that (and first of all
> > prove or disprove my guess).
> Guess proved!
> Attached find extensions to the messaging-fdpass tests
> that prove it and add regression safety in that respect:
> I add two tests
> These are variants of the LOCAL-MESSAGING-FDPASS2
> test that sent a very small payload (FDPASS2a)
> and no payload at all (FDPASS2b) with the message.
> Both trigger the fastpath message sending.
> Without Noel's patch both tests hang.
> With Noel's patch they pass.
> Review/push appreciated.
Pushed after getting review from Guenther on a side-channel.
Thanks - Michael
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 198 bytes
Desc: not available
More information about the samba-technical