Patches suck time (Greedy little performance enhancing vermin that they are...)


So, other than spending... I don't even want to think about how much time... working with D'Arcy this morning, I worked on getting the patch for PySNMP put together (SourceForge's CVS being back online). That was going to be a 20 minute task before I started the real work on fixing the OID object mechanism...

Of course, I had a lot of changes in my PySNMP tree that really weren't ready for deployment, or which shouldn't be part of the patch, they were just stuff I was trying at some point and never got around to reversing. So I spent quite a bit of time eliminating those from the patch.

Then I realised I had no idea whether the new patch worked, so I spent a good bit of time getting a new checkout set up and the patch applied to it... that went fine once I remembered that I am working against the maintenance branch, not the head...

Then I realised I hadn't done any timings to measure total effect, so couldn't explain why I wanted it... so I had to check out yet another version and run timing tests. Spent a good long while running test suites (and getting confused by results) before I realised I'd turned off the profiling in the TwistedSNMP test runner (so the stats were not getting re-generated).

Finally got the patch in and decided to do the deployment on the live system... spent a while trying to figure out why that patch couldn't find files... then remembered the branch (duh!) and settled down to patch and restart... only to discover the flipping patch, which applies perfectly well on the Win32 machine, fails for just about every chunk on the FreeBSD box.

At which point, having been working since an unreasonable hour in the morning, I decided to stop and give my heart a rest (stress does bad things to your body, especially when you get basically no exercise in the course of a normal day).

Comments

Comments are closed.

Pingbacks

Pingbacks are closed.