Comment 10 for bug 405317

I would be surprised if it was machine IO related since it occurs on many machines (Linux and Windows) suddenly since 1.16, and other IO intensive software (Eclipse on that same ws) is as fast as usual. Trashing would be unlikely as the machine for that report has 6GB memory of which about 3GB were free at the time. I also have a desktop memory/CPU load thingy running all the time and lots of time to look at it while commiting ;-) and it did not do anything special while commiting wrt memory (CPU was up because 1/8 CPU was on 100%).

Is there anything common in the code for pull and commit? Since pull is horrible too.. I cannot really imagine but it is remarkable that both local and remove stuff breaks in the same way in the same version.

I will do the commits as you specified; I assume the syntax would be:
bzr --lsprof-file commit.callgrind commit -m "foo"
instead of the double commit? I will try this tomorrow (1:30am here now) asap. Thanks a lot for your help!