On Fri, 08 Apr 2011 01:53:01 +0530, Greg Freemyer <greg.freemyer@gmail.com> wrote:

I don't think bisecting lets you do that.

You give it a known good git revision and a known bad.

It picks a revision 50% between those and compiles it.

You test it and tell git if is good or bad. It then gets a revision 50% between the new known good and known bad pair and repeats.

So its a log2 algorithm at finding the checkin that broke things.

So 1 million checkins take only 20 compile/test cycles.

ie. 2 **20 = (2 ** 10) ** 2 = 1024 **2 = 1024 * 1024 ~= 1,000,000

Thus with a huge number of checkins, it is relatively fast to get to the answer.

The trouble is with only 500 checkins, it still takes 9 compile / test pairs.

Not so good, but still a lot better than doing all 500.

thank you. not only haven't i done it, i didn't even understand it properly. good to know. and, in case of applications of the size of libreOffice, quite tiresome. -- phani. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org