On 14.07.2013 23:40, Rainer Jung wrote: > On 14.07.2013 23:05, Rainer Jung wrote: >> On 11.07.2013 20:54, Jim Jagielski wrote: >>> The pre-release test tarballs for Apache httpd 2.4.5 can be found >>> at the usual place: >>> >>> http://httpd.apache.org/dev/dist/ >>> >>> I'm calling a VOTE on releasing these as Apache httpd 2.4.5 GA. >>> NOTE: The -deps tarballs are included here *only* to make life >>> easier for the tester. They will not be, and are not, part >>> of the official release. >>> >>> [ ] +1: Good to go >>> [ ] +0: meh >>> [X] -1: Danger Will Robinson. And why. > > I have to revert my +1. The patch applied for PR54948 breaks vhost mapping. > > It is > > http://svn.apache.org/viewvc?view=revision&revision=1485675 > > backported from > > http://svn.apache.org/viewvc?view=revision&revision=1481306 > > The change was not applied to 2.2. > > The config in question has a default vhost plus a wildcard vhost plus > lots of specific vhosts (all for the same 443 port) and version 2.4.5 > only shows the default and the wild card vhost in the "-S" output and > the config works the same way, so the more specific vhosts are hidden. > > Using 2.4.4 or undoing the single commit for 2.4.5 shows the expected > behavior. > > I need to create a more simple config for reproduction and follow what > happens in the above code change.
I propose the following patch: http://people.apache.org/~rjung/patches/vhost-pr54948-part2.patch Caution: I did not really understand that code, but tracked what happened during digesting the broken config using additional log output. The original patch for PR54948 not only removed the unwanted internal duplicates but also dropped the 443 part from any ":80 :443" VirtualHost. Someone knowing this code better should confirm, whether my addition is correct or whether PR54948 should be fixed in a different way. IMHO the current 2.4.5 code is really broken and we should either release the code with r1485675 reverted or an additional fix on top. The config that was broken is our ASF www.apache.org config. Version 2.4.5 ignored the 443 part of most of the ":80 :443" vhosts, more precisely all except for the default vhost and the first internally processed one. Since the first processed one was the last declared one, which was originally meant as a fall through catch all, that vhost now handled all 443 traffic. Regards, Rainer P.S.: It is getting late here, so I will unfortunately soon go offline.
