EXPERIMENTAL: tevent_kqueue support

[Jeremy Allison]

On Wed, Feb 27, 2013 at 06:39:15PM +0100, Stefan (metze) Metzmacher wrote:
> Am 27.02.2013 17:58, schrieb Jeremy Allison:
> > On Wed, Feb 27, 2013 at 07:02:43AM +0100, Stefan (metze) Metzmacher wrote:
> >> Am 26.02.2013 22:44, schrieb Jeremy Allison:
> >>> On Tue, Feb 26, 2013 at 01:35:58PM -0800, Jeremy Allison wrote:
> >>>>
> >>>> That code was already written carefully to do exactly what
> >>>> it does :-).
> >>>>
> >>>> Without using threads it's really easy to get yourself
> >>>> blocked on a pipe. I'll have to do more testing to see
> >>>> if it's true that writable is cleared when a pipe is
> >>>> full.
> >>>
> >>> Yep. Just confirmed with junk code. poll returns
> >>> writable on a pipe fd even if writing to such a
> >>> pipe would block.
> >>>
> >>> That means that for this test we can't always
> >>> prioritize writes over reads, as we just end
> >>> up blocked in the callback. That's why the
> >>> callback code is written as it is, I'm
> >>> assuming a pipe can hold 256 btyes without
> >>> blocking.
> >>
> >> If we get POLLOUT without being able to write 1 byte
> >> without blocking, we have a problem.
> >> That's the whole point of having POLLOUT at all...
> >>
> >> Can you post your junkcode example?
> > 
> > As I mentioned later, I had a bug :-).
> > 
> > Got it working now. I'll re-do the test patch
> > to work without the hack and re-post it.
> > 
> > You can then rebase it on top of the merged
> > patchset.
> 
> I have a specific test for the full buffer thing now...
> 
> https://gitweb.samba.org/?p=metze/samba/wip.git;a=commitdiff;h=dbbf42e848d2d125a2d35916bc2482a51ee3994b
> 
> It generates the attached strace output, which shows that all backends
> behave the same.

Except they don't quite :-).

I found a bug in the tevent select backend, using the attached patch.

It turns out that when we're deciding what handlers to call
in the select backend, we don't take into account the fact
that the same fd may have been added into the read FD_SET
and the write FD_SET but with different handlers.

We just match on the file descriptor, and we must match
on both the file descriptor and the flags requested before
calling the handler.

Attached are the select bugfix, and the fixed regression
test for the multiple fd case.

Without the select bugfix the regression test for the
multiple fd case panics when fde_handler_read_1() is
called (which should never happen).

So I think we probably could do with both my test
and your test in order to cover all cases over all
backends.

Cheers,

	Jeremy.
fd
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Fix-multiple-handler-on-the-same-fd-bug-in-the-teven.patch
Type: text/x-diff
Size: 1481 bytes
Desc: not available
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20130227/7556e2c6/attachment.patch>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Regression-test-to-ensure-that-a-tevent-backend-can-.patch
Type: text/x-diff
Size: 4497 bytes
Desc: not available
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20130227/7556e2c6/attachment-0001.patch>