Hi. We're using bzr on several tiny projects, each with fewer than 5000 files and less than 100MB of store. So far we have not deployed bzr for any big projects -- the ones that pay the bills -- and I've posted some of my observations here in the past. The recent posts on this list about possible future speed and memory optimizations sound promising. So I'm getting ready to update my testing to 2.2 so that I'll have a fresh baseline for comparison whenever improvements are available for testing. As usual, I have found something interesting other that what I was expecting. My question may span multiple versions of qbzr, but for completeness I am running bzr 2.2b4 with qbzr 0.19b2 and explorer 1.1b1. Python is 2.6.4. First, some context for my questions: I have just created a new shared repo (format 2a) and populated a trunk branch with a project of 15,000 files, including some of binaries (bitmap images and closed-source vendor libs). Total size is 1.1 GB. It took about six minutes ( 357 seconds) to commit the files to trunk. I have added a few ignores but mostly it's the usual stuff. Here comes my main observation: I branch from trunk to create a branch dev_a in the same shared repo. From the time I press the OK button in the qbranch dialog to the time the content appears complete and I am presented with a Close button is 4.75 minutes. When I press Close, the qbranch dialog does not close. Instead, it goes off an does something (?) for an additional 1.5 minutes, during which time the Close Button is grayed out. After the 1.5 minute delay, the Close button is re-enabled and clicking it immediately closes the qbranch dialog. Is this normal? Why do I have to click Close twice? And what is qbranch doing during the 1.5 minute apparent pause? Note that everything above was done in explorer 1.1b1. If I instead use the commandline 'bzr qbranch trunk dev_b' to create another branch in the shared repo, the time from pressing OK to seeing the Close button appear is just under 5 minutes (4:58) and the qbranch window closes all by itself. I consider the difference between 5.0 minutes and 4.75 minutes to be irrelevant for the moment, but the difference between 4.75 + 1.5 = 6.25 minutes and 5.0 minutes is not irrelevant. So I'm still really wondering what qbranch is doing differently when invoked from explorer than when invoked from the command line. Finally, just to round things out, I ran 'bzr branch trunk dev_c --use-existing-dir" to see how bzr would do all by itself: four minutes and ten seconds. So it seems that in addition to the difference between qbranch under explorer and qbranch from the commandline, it may also be true that qbranch adds a fair bit of overhead compared to branch. What could that be? Is the Qt event loop really that greedy? Thanks ~M