Commit eec276d7 authored by Jim Jagielski's avatar Jim Jagielski
Browse files

promote 


git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/branches/2.4.x@1674920 13f79535-47bb-0310-9956-ffa450edef68
parent b850570d

STATUS

+30 −30
Changes for STATUS: 30 added lines, 30 removed lines.
Original line number Diff line number Diff line
@@ -105,6 +105,36 @@ RELEASE SHOWSTOPPERS: 
PATCHES ACCEPTED TO BACKPORT FROM TRUNK:

  [ start all new proposals below, under PATCHES PROPOSED. ]



  *) mpm_event: Allow for timer events duplicates.

     trunk patch: http://svn.apache.org/r1666468

                  http://svn.apache.org/r1666618

                  http://svn.apache.org/r1674697

     2.4.x patch: http://people.apache.org/~ylavic/httpd-2.4.x-event_timers_duplicates-v2.patch

     +1: ylavic, covener, jim

     covener: what's the background here? What's currently broken?

     ylavic: I expected to find a common compare function for both MPMs event

             and motorz, but I now think that won't be the case (not the same

             constraints for both since event does not apr_skiplist_remove(),

             see http://marc.info/?l=apache-httpd-dev&m=142714857911558&w=2 for

             details). Hence I'm moving this backport back to proposals, it is

             IMO the right fix to not lose timers registered in the same micro-

             second (in MPM event).

     jim:    As I see it, the current code in Event does work as expected. The

             above simply adds some checks, does it not? As far as the logic

             is concerned, what exists and what is proposed does the exact

             same thing, doesn't it?

     covener: IIUC, currently event cannot insert two timer_event_t's that would fire at the

              same apr_time_t because the compare function would return 0 and insert_compare

              wouldn't let the duplicate in. post-patch, only two timer_event_t's with the same

              address are equals.

     ylavic: r1674697 (and v2) removes the debugging checks (not necessary in 2.4.x IMHO),

             and add a comment about the compare function. As Eric explained above, the real

             point is to return +1 instead of 0 there, so that duplicates are added after

             each other (by order of insertion).

        jim: OK, perfect. Since 2.4 uses apr-1.5, we need to use the same

             logic-flow as the comp() function in testskiplist to ensure

             that (1) dups are allowed and (2) inserted in the correct place.





PATCHES PROPOSED TO BACKPORT FROM TRUNK:

  [ New proposals should be added at the end of the list ]
@@ -274,36 +304,6 @@ PATCHES PROPOSED TO BACKPORT FROM TRUNK: 
     2.4.x patch: trunk works (modulo CHANGES)

     +1: ylavic



  *) mpm_event: Allow for timer events duplicates.

     trunk patch: http://svn.apache.org/r1666468

                  http://svn.apache.org/r1666618

                  http://svn.apache.org/r1674697

     2.4.x patch: http://people.apache.org/~ylavic/httpd-2.4.x-event_timers_duplicates-v2.patch

     +1: ylavic, covener, jim

     covener: what's the background here? What's currently broken?

     ylavic: I expected to find a common compare function for both MPMs event

             and motorz, but I now think that won't be the case (not the same

             constraints for both since event does not apr_skiplist_remove(),

             see http://marc.info/?l=apache-httpd-dev&m=142714857911558&w=2 for

             details). Hence I'm moving this backport back to proposals, it is

             IMO the right fix to not lose timers registered in the same micro-

             second (in MPM event).

     jim:    As I see it, the current code in Event does work as expected. The

             above simply adds some checks, does it not? As far as the logic

             is concerned, what exists and what is proposed does the exact

             same thing, doesn't it?

     covener: IIUC, currently event cannot insert two timer_event_t's that would fire at the

              same apr_time_t because the compare function would return 0 and insert_compare

              wouldn't let the duplicate in. post-patch, only two timer_event_t's with the same

              address are equals.

     ylavic: r1674697 (and v2) removes the debugging checks (not necessary in 2.4.x IMHO),

             and add a comment about the compare function. As Eric explained above, the real

             point is to return +1 instead of 0 there, so that duplicates are added after

             each other (by order of insertion).

        jim: OK, perfect. Since 2.4 uses apr-1.5, we need to use the same

             logic-flow as the comp() function in testskiplist to ensure

             that (1) dups are allowed and (2) inserted in the correct place.



  *) mod_authz_core: Allow "Require expr" to work when the expression is quoted,

     as in an example in the doc.  PR 56235

     trunk patch: http://svn.apache.org/r1585609