I updated lighttpd to this version less than 24 hours ago, and it has already crashed three times, seemingly in mod_fastcgi. The previous version worked fine. Running this in CentOS 7.2.1511, to which I updated some 15 days ago. I'm still investigating the exact reason for the crashes.
The EPEL package ships with mod_geoip, which is not included in stock lighttpd. I believe lighttpd-mod_geoip.c would need to be updated for 1.4.38, because the current config_insert_values_global function call takes four arguments instead of three. All the modules included in stock lighttpd were modified because of this change. I'm not using mod_geoip myself.
I'm going to give this a -1 on the grounds that 1.4.38 crashed a dozen times. For the first two crashes I had to restart lighttpd manually, causing some 2.5 hours of downtime before I managed to restart lighttpd. For the latter crashes I had scripted lighttpd to be restarted automatically. I and others provided some debugging data, which ended up with a patch as mentioned in https://redmine.lighttpd.net/issues/2700#note-6
I have been running the patched version for two days now with zero crashes, so looks like the patch works. I would recommend applying the patch for all the EPEL and Fedora builds of lighttpd. Reproducing this bug on quiet development servers seems to be difficult, I only ran into these problems on a production web server. I also have two web servers running the non-patched EPEL6 1.4.38 version, but those don't seem to have enough traffic to trigger this bug. My EL7 webserver serves 1000x more requests.
This update has been submitted for testing by limb.
This update has been pushed to testing.
I updated lighttpd to this version less than 24 hours ago, and it has already crashed three times, seemingly in mod_fastcgi. The previous version worked fine. Running this in CentOS 7.2.1511, to which I updated some 15 days ago. I'm still investigating the exact reason for the crashes.
https://redmine.lighttpd.net/issues/2700 is related, although not filed by me.
The EPEL package ships with mod_geoip, which is not included in stock lighttpd. I believe lighttpd-mod_geoip.c would need to be updated for 1.4.38, because the current config_insert_values_global function call takes four arguments instead of three. All the modules included in stock lighttpd were modified because of this change. I'm not using mod_geoip myself.
Oh, scratch my previous comment re mod_geoip, this seems to be taken care of already.
I'm going to give this a -1 on the grounds that 1.4.38 crashed a dozen times. For the first two crashes I had to restart lighttpd manually, causing some 2.5 hours of downtime before I managed to restart lighttpd. For the latter crashes I had scripted lighttpd to be restarted automatically. I and others provided some debugging data, which ended up with a patch as mentioned in https://redmine.lighttpd.net/issues/2700#note-6
I have been running the patched version for two days now with zero crashes, so looks like the patch works. I would recommend applying the patch for all the EPEL and Fedora builds of lighttpd. Reproducing this bug on quiet development servers seems to be difficult, I only ran into these problems on a production web server. I also have two web servers running the non-patched EPEL6 1.4.38 version, but those don't seem to have enough traffic to trigger this bug. My EL7 webserver serves 1000x more requests.
This update has reached 14 days in testing and can be pushed to stable now if the maintainer wishes
This update has been submitted for stable by limb.
This update has been pushed to stable.
IMHO sending the update with -1 karma to stable without further communication is not a good idea. This update broke lighttpd on EPEL5 as well.