On Tuesday 31 January 2012, Rainer Jung wrote: > On 30.01.2012 21:11, Jim Jagielski wrote: > > Do I dare float the idea of a 2.4.1 T&R "very soon"?? > > I'm done with flodding the list with trivial backports. > > Still open are: > > A) Testing/Fixing the "bucket problems". There were two bugs, and I > have lost the overview, how the fixes are related. One was the > problem noted under Windows when SSL was failing due to an early > read failure (BZ 5476), one was the problem noted by Joe about > reading whole CGI/PIPE-like mutating bucket types into RAM.
The CGI/PIPE-like mutating bucket bug should now be completely fixed in 2.4, as I have just backported r1234899. r1233882 should probably be backported, too, but is unrelated. The other changes (including my patch proposed on the list) are just fixing a broken API but should not result in observable behavior change. I don't think that they will help with the Windows/SSL issue. > There is the patch proposed by Stefan for trunk. For 2.4.x there > are some backports missing in addition. I guess all of the > following would be needed for 2.4.x (plus Stefan's patch): > > ------------------------------------------------------------------- > ----- r1236122 | rpluem | 2012-01-26 11:03:36 +0100 (Thu, 26 Jan > 2012) | 1 line > > * Don't typedef twice (in .c and .h file). > ------------------------------------------------------------------- > ----- r1233882 | jorton | 2012-01-20 13:41:18 +0100 (Fri, 20 Jan > 2012) | 4 lines > > * server/core_filters.c (ap_core_input_filter): Only treat EAGAIN > as success if a non-blocking read was requested; for a blocking > read, it is an error condition. > > ------------------------------------------------------------------- > ----- r1234899 | jorton | 2012-01-23 17:57:07 +0100 (Mon, 23 Jan > 2012) | 4 lines > > * server/core_filters.c (send_brigade_nonblocking): Use a > non-blocking bucket read, allowing any pending data to be flushed > before trying a (potentially slow) blocking read. > > ------------------------------------------------------------------- > ----- r1235019 | sf | 2012-01-23 22:58:42 +0100 (Mon, 23 Jan 2012) > | 6 lines > > Make the core input/output filter contexts private and provide > accessor APIs for mpm_winnt and mod_ftp. > > This allows to add members to the context structs without breaking > binary compatibility. > > ------------------------------------------------------------------- > ----- > > > B) further open stuff from STATUS: > > * Docs about building / installing are not up to date. > > * PR 52402: balancer crash on Windows > > > C) some of the points in the separate "Questions" mail thread. I > don't see a shot stopper there, but some points should be easy: > > > 1) mod_reqtimeout: not a showstopper, but would be good to have > consensus about preferred behaviour so we do not need change > defaults dramatically after GA. > > 2) log tags: I hope Stefan will tell us, whether we should backport > the make targets and the tags directory in docs/. > > 3) Recent install optimization: again I would rely on Stefan > whether he thinks the changes are safe. It is only about the build > system, so could be backported after GA. I have reverted 1) in trunk. Only a small part of 2) should go into 2.4, but that can wait. 3) should see some more testing before backporting. > 5) mod_authnz_ldap: I hope Eric can comment on that one. > > 8) ErrorLog directory checking: Could be backported later, but > changes behaviour under error conditions, so maybe better to > include now unless Stefan thinks it is a risk. I think Stefan was > aksing for Windows testers here. > > 4) and 7) are already done, 6) [cache] seems to be consensus for > not changing and 9) [event] seems to be OK as is. 8) Is now done, too. This leaves only 5) from Rainer's list.
