Archives week 52 of 2004
Dec. 27, 2004 - Jan. 2, 2005
This section discusses the use of play in the sense of loosness, in the design process. It's obviously going to need a lot more text, as the number of issues it touches is large. Still, the basic idea is there. Could use a lot more playfulness in the writing, I think. Anyway, read away, ...
Making the pointer handling code simpler is proving rather involved. PyOpenGL is written (not by me, this is all Tarn's code) as a set of macros that expand 4 or 5 levels deep even before it gets to SWIG, which is essentially an interlocking set of macro expansions. The problem is that the pointer-handling basically ...
I'm planning on giving my camera to Shane when I buy a replacement I can actually use. Was intending to get out this week to buy it, but instead I wound up spending the entire time working.
Shane reminded me about it again last night and suggested I try Shopper's drug mart, as they are ...!-->!-->
Was just sitting down to start work on PyOpenGL when Shane and Simon called from Lua's party asking me to bring along my Cranium game. So, toddled down to her place (she lives about 4 blocks away) and spent the evening playing Cranium (well, we didn't even finish a single game) then sitting around discussing ...
Spent the evening after work on trying to track down why there's a memory leak in the pointer-management functions in PyOpenGL. Problem is, the code there is this grotty low-level stuff that winds up re-creating the reference-counting mechanism in Python just to track whether anything is referencing a pointer.
Decided to rework it as a ...!-->!-->
Other than some brief over-the-phone debugging, mostly worked on little issues all day, adding links to various pages, refactoring things, adding paging to a long table views, the kind of cleanup you need to do before a project becomes a product.
Still quite a bit to do that way, but that's tomorrow's bugbear. Tonight I'm ...!-->!-->
Only changes are to the BisectOIDStore; allowing it to have sub-OIDStores registered for handling a given table, and allowing it to have callable objects registered to return dynamically calculated values for a given OID.
Those were just features I needed for testing Cinemon, but they should make the writing of test code far simpler in ...!-->!-->
TwistedSNMP is fairly minimal in what it allows on the Agent side, basically static tables mapping OID to value. That's been fine up to now, but I realised as I was trying to track down a bug in Cinemon this afternoon that the reason all of my modems are showing 0-values for their counters in ...
Spent the bulk of the day finishing and then debugging the new hierarchy type, and then figuring out why an old hierarchy type seemed to have been messed up (it wasn't, it's always worked like that, I just never noticed it).
After that, went to our cousin's husband's sister's house for dinner and a game ...!-->!-->