[Bug 4662] New: `hg bisect -s` should not use linear probing to choose the next revision

[mercurial-bugs at selenic.com]

http://bz.selenic.com/show_bug.cgi?id=4662

          Priority: normal
            Bug ID: 4662
                CC: mercurial-devel at selenic.com
          Assignee: bugzilla at selenic.com
           Summary: `hg bisect -s` should not use linear probing to choose
                    the next revision
          Severity: bug
    Classification: Unclassified
                OS: Mac OS
          Reporter: n.nethercote at gmail.com
          Hardware: PC
            Status: UNCONFIRMED
           Version: unspecified
         Component: Mercurial
           Product: Mercurial

I just bisected a regression in Firefox. There were a couple of longish
sequences of revisions where there were compile errors, so I used `hg bisect
-s`. Unfortunately, `hg bisect -s` chooses the next revision using (AFAICT)
linear probing, going forward or backward one revision at a time. This is very
slow and I ended up doing part of the bisection manually, which was much
faster. 

Eventually, I manually bisected to one side of the broken sequence and `hg
bisect` was able to behave usefully again.

This is a problem I've encountered multiple times now. It would be great if `hg
bisect -s` did something smarter -- something less linear and more log2-ish.
Thank you.

-- 
You are receiving this mail because:
You are on the CC list for the bug.