RFC: startup time - again

[Mark Hammond]

Hi Alex, thanks for the info.  My times are:

Average of 10 runs with original .zip: 0.211, best was 0.197
10 runs without bzrlib\test: 0.195, best was 0.180

I did a little experimenting with compression levels etc, which didn't change the numbers much, then tried again with brand-new .zip files

Average of 10 runs with original .zip: 0.205, best was 0.198
10 runs without bzrlib\test: 0.201, best was 0.199

So while there was a little improvement, it still seems very close to being within noise levels.

Theoretically, playing with the .zip compression level should change the results depending on your hardware - eg, a fast processor and slow disk would theoretically benefit from high compression.  py2exe defaults to using no compression IIRC.  But as mentioned, I couldn't get any significant differences at any compression level.  Another interesting option would be to try and split out the tests into their own .zip file and add that at the end of sys.path (and it might end up making sense to do something similar for all plugins?)

Cheers,

Mark