As reported in issue 1445:
A valid candidate revision for a bisect test is not considered for testing
due to its skipped ancestor. If this revision is the only untested one left
an assertion error occurs.
The real reason for both issue is that bisect can not handle cases where there
are multiple possibilities for the result.
Example (from issue1228):
rev 0 -> good
rev 1 -> skipped
rev 2 -> skipped
rev 3 -> skipped
rev 4 -> bad
Note that this patch does not only fix the reported Assertion Error but also
the problem of a non converging bisect:
hg init
for i in `seq 3`; do echo $i > $i; hg add $i; hg ci -m$i; done
hg bisect -b 2
hg bisect -g 0
hg bisect -s
From this state on, you can:
a) mark as bad forever (non converging!)
b) mark as good to get an inconsistent state
c) skip for the Assertion Error
Minor description and code edits by pmezard.
Automatically detect whether we're looking for a bad to good
transition rather than the usual good to bad transition by detecting
when badrev is inside the good set and flipping good/bad.