gtkmm API freeze

Just lost another weekend to gtkmm. Andreas Holzmann fixed our last API bug, so I thought we were ready for a source-frozen release. But then Michael v. Szombathely found that we were leaking GtkDialogs. It turned out to need quite a hefty recoding of the various bits of code that link the gtkmm and GTK+ object lifecycles. I spent all of Saturday and most of Sunday doing that, and the code is now even better documented than before.

But I now find that one or two of our examples, particularly the menus, segfault on quit, somewhere in memcpy(). I just downloaded Valgrind and it looks like it's going to be very helpful. I'm already impressed.

I'm pleased to see that julian is working on the gnomemm2 port of Gabber. Like everyone he's finding that you have to completely recode for TreeView. But it could be worse – he could be doing it with the C API. I'm glad that I created a generic grid interface for Glom. It's still a lot of work to convert from GtkExtra::Sheet to Gtk::TreeView, but at least it's all in one place.

The past 2 weeks have been hellish at work. I've been fighting with MSMQ and it's apparent inability to either send or receive messages. I've never known any piece of software to so defy prediction. No two tests ever yield the same results. In a sane world I'd be able to quickly see what was wrong by looking at the code. Unfortunately I can't seem to find any open-source message queues that don't have only Java interfaces.

I went to the South Bank's Hayward gallery the other day, to see photography exhibitions by William Eggleston and Ansel Adams. I particularly recommend the first one.