Henry Jen wrote:
As the new branches got create, I am wondering if anyone can have a look
at this patch.
I agree the pollset patches should go in. Its partly my fault for not
getting to them earlier... but if someone else has time, please
review+commit them.
Also there is a patch for thread pool at bugzilla[1] I think is worth to
commit.
[1] https://issues.apache.org/bugzilla/show_bug.cgi?id=43876
However, this one seems to mix new APIs with bug fixes/race conditions,
which makes it very hard ti handle imo.
-Paul
---------- Forwarded message ----------
From: *Henry Jen* <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>>
Date: Mon, Dec 3, 2007 at 11:22 PM
Subject: [PATCH] Solaris port fix (bug 43000)
To: APR Developer List <[email protected] <mailto:[email protected]>>
Hi,
It's holiday season and things may be moving slower than it usually
is. The patch have been available for a while, and it fixes the port
code for Solaris which I really like to get it in.
It could be somewhat controversial for the atomic code used to defer
the port_associate in apr_pollset_add in order to pass the testpoll
code, which is the needed fix if we consider the semantic of
apr_pollset_poll should be giving accumulated events; Otherwise, we
should revise the test/testpoll.c.
With that said, I break the patch into two, the first port_1203-1.diff
fix bugs of:
1. Return APR_SUCCESS if the poll removed from pollset is in the
add_ring, which means the poll had been added but is current
disassociated because an event.
2. In apr_pollcb_poll, call apr_pollcb_add to ensure port_associate.
And the port_1203-2.diff is the fix to pass testpoll given current
situation, which is less a concern as I consider in most apps,
apr_pollset_poll is called in a loop.
Any comments are welcomed.
Cheers,
Henry
