[development] open questions
Hi, I had some distraction in the last two weeks, thus I was unable to follow the syslog-ng mailing list/emails as closely I would have wanted. I've finished reading all the unread emails in my syslog-ng folder, and I wanted to summarize the status quo: Bugfixes/maintenance (Bazsi): * I've merged 3.2 into 3.3 to get the latest changes, bugfixes. libsyslog-ng-patterndb.so (Corinna, Marius Tomaschewski, Peter Czanik): * causes problems on cygwin & SUSE, so probably an alternative solution is needed. * my current idea is either to use Corinna's idea and put pdbtool in /usr/lib/syslog-ng (which I personally don't really like, as /usr/lib is not the place for binaries), or that libsyslog-ng-patterndb.so should be moved /usr/lib _after_ compilation by the distribution build scripts. * Marius, Corinna what's your take on this issue? $(format-json) template function (Balint, Algernon, Matthew) * waiting for the generic value-pairs() implementation * it could probably be added to 3.2 as well, if value-pairs doesn't prove to be risky in itself (or else they'd have to be put in the tfjson module). mongodb dynamic values support (Algernon) * waiting for the value-pairs() implementation (Algernon) * this is 3.3 stuff systemd support (Marius Tomaschewski, Tom Gundersen, Algernon) * should be more-or-less ready, Algernon promised to rebase it against both 3.2 and 3.3 , so that I can integrate into both. LogQueue scalability (Bazsi): * I was working on making LogQueue more scalable -- Bazsi
$(format-json) template function (Balint, Algernon, Matthew)
* waiting for the generic value-pairs() implementation
* it could probably be added to 3.2 as well, if value-pairs doesn't prove to be risky in itself (or else they'd have to be put in the tfjson module).
Matthew backported the non-value-pairs() version to 3.2 already. I'm not sure whether it's worth adding value-pairs() to 3.2 - even if it's reasonably easy and self-contained code, it's still seems too big for 3.2. On the other hand, if it's deemed worthy enough, I'll happily backport both value-pairs() and tfjson() to 3.2. However, before all that, value-pairs() needs to be finished. I plan to write a status report tonight.
systemd support (Marius Tomaschewski, Tom Gundersen, Algernon)
* should be more-or-less ready, Algernon promised to rebase it against both 3.2 and 3.3 , so that I can integrate into both.
This is available from the integration/systemd branches of my 3.2 and 3.3 trees: git://git.madhouse-project.org/syslog-ng/syslog-ng-3.2.git and git://git.madhouse-project.org/syslog-ng/syslog-ng-3.3.git The integration branch was based on the current master of of each version, and contains two commits: one that makes GSockAddr->sa_funcs per-instance instead of per-family. On top of this is Marius' systemd socket activation support (condensed into a single commit, including my inherit_systemd_activation() cleanup). -- |8]
systemd support (Marius Tomaschewski, Tom Gundersen, Algernon)
* should be more-or-less ready, Algernon promised to rebase it against both 3.2 and 3.3 , so that I can integrate into both.
This is available from the integration/systemd branches of my 3.2 and 3.3 trees:
git://git.madhouse-project.org/syslog-ng/syslog-ng-3.2.git and git://git.madhouse-project.org/syslog-ng/syslog-ng-3.3.git
The integration branch was based on the current master of of each version, and contains two commits: one that makes GSockAddr->sa_funcs per-instance instead of per-family.
On top of this is Marius' systemd socket activation support (condensed into a single commit, including my inherit_systemd_activation() cleanup).
Tom Gundersen contributed an example syslog-ng.service file, which is available from the integration/systemd-with-example branch (for both 3.2 and 3.3) from the repositories above. I was contemplating merging it to the main integration/systemd branch, but for reasons I already forgot I opted for a separate branch instead. Bazsi: please pull whichever you prefer. :) -- |8]
On Sun, 2011-02-06 at 18:13 +0100, Gergely Nagy wrote:
$(format-json) template function (Balint, Algernon, Matthew)
* waiting for the generic value-pairs() implementation
* it could probably be added to 3.2 as well, if value-pairs doesn't prove to be risky in itself (or else they'd have to be put in the tfjson module).
Matthew backported the non-value-pairs() version to 3.2 already. I'm not sure whether it's worth adding value-pairs() to 3.2 - even if it's reasonably easy and self-contained code, it's still seems too big for 3.2.
On the other hand, if it's deemed worthy enough, I'll happily backport both value-pairs() and tfjson() to 3.2.
However, before all that, value-pairs() needs to be finished. I plan to write a status report tonight.
I'll have a look and decide based on that.
systemd support (Marius Tomaschewski, Tom Gundersen, Algernon)
* should be more-or-less ready, Algernon promised to rebase it against both 3.2 and 3.3 , so that I can integrate into both.
This is available from the integration/systemd branches of my 3.2 and 3.3 trees:
git://git.madhouse-project.org/syslog-ng/syslog-ng-3.2.git and git://git.madhouse-project.org/syslog-ng/syslog-ng-3.3.git
The integration branch was based on the current master of of each version, and contains two commits: one that makes GSockAddr->sa_funcs per-instance instead of per-family.
On top of this is Marius' systemd socket activation support (condensed into a single commit, including my inherit_systemd_activation() cleanup).
I have now pulled your latest patches into both 3.2 and 3.3 (the one which implements my afsocket_acquire_socket() virtual method suggestion), so now both syslog-ng 3.2 & 3.3 sports systemd support. Great work, thanks. -- Bazsi
participants (2)
-
Balazs Scheidler
-
Gergely Nagy