On Tue, Jan 8, 2013, at 10:02 PM, Sébastien Michel wrote: > 2013/1/3 Bron Gondwana <br...@fastmail.fm>: > > > > I vote for using the technically best library. It's probably small enough > > to that > > we can ship it inside the Cyrus tarballs if we need. It's only 132k in the > > src > > directory of the latest jannson, and it's MIT licensed. > > 2013/1/4 Greg Banks <g...@fastmail.fm>: > > > > As long as the feature gracefully disables itself if the library is not > > installed at configure time, I'm really not fussed which library we use. > > For the CI server we'll probably need to build it manually, like a > > whole bunch of other things. > > > > If we choose to include the latest jansson sources in the Cyrus > tarbal, It should be only if the jansson library is missing on the > host server. > In this case event notification may be always compiled and it will be > possible to make some cleanup : > - removing test for event-notification in configure.ac > - removing ENABLE_MBOXEVENT C macro in some source files (imapd.c, > pop3d.c, etc.)
Those can be made cleaner anyway, by moving them into mboxevent.c. Then the mailbox.c code would always call mboxevent_* functions but those functions would, depending on the result of ./configure, either be a trivial stub that does do nothing or would actually do the work of pushing the event out. > - removing MBOXEVENT macro in Makefile.am In the scenario above you would always compile mboxevent.c, so this automake conditional would not be needed. -- Greg.
