[issue1953] incomplete bisection in 1.4 and later

[John Coomes]

New submission from John Coomes <john.coomes at sun.com>:

hg bisect reports the wrong changeset as the 'first bad revision', stopping
testing too early.  Use the attached script ('repro') to reproduce.  It
clones a moderate-sized public repo; I was unable to reproduce the problem
in a small repo which mimics the branch structure of the public one.

Running the repro script w/hg 1.4.1, the bisect result is incorrect:

requesting all changes
adding changesets
adding manifests
adding file changes
added 1053 changesets with 9547 changes to 3856 files
updating to branch default
3659 files updated, 0 files merged, 0 files removed, 0 files unresolved
Mercurial Distributed SCM (version 1.4.1)
Testing changeset 1029:b0b36f0de97e (8 changesets remaining, ~3 tests)
42 files updated, 0 files merged, 1 files removed, 0 files unresolved
Changeset 1029:b0b36f0de97e: bad
bool ParallelScavengeHeap::is_in_young(oop* p) {
Changeset 1017:e715b51789d8: good
bool ParallelScavengeHeap::is_in_young(oop* p) {
Changeset 1026:f99f695bb8ef: good
The first bad revision is:
changeset:   1029:b0b36f0de97e
parent:      1028:052a899eec3e
parent:      1026:f99f695bb8ef
user:        tonyp
date:        Tue Oct 20 11:28:25 2009 -0700
summary:     Merge

However, hg 1.3.1 finds the correct changeset:

requesting all changes
adding changesets
adding manifests
adding file changes
added 1053 changesets with 9547 changes to 3856 files
updating working directory
3659 files updated, 0 files merged, 0 files removed, 0 files unresolved
Mercurial Distributed SCM (version 1.3.1)
Testing changeset 1017:e715b51789d8 (31 changesets remaining, ~4 tests)
67 files updated, 0 files merged, 1 files removed, 0 files unresolved
bool ParallelScavengeHeap::is_in_young(oop* p) {
Changeset 1017:e715b51789d8: good
bool ParallelScavengeHeap::is_in_young(oop* p) {
Changeset 1025:1ee412f7fec9: good
Changeset 1029:b0b36f0de97e: bad
Changeset 1028:052a899eec3e: bad
Changeset 1027:39b01ab7035a: bad
The first bad revision is:
changeset:   1027:39b01ab7035a
parent:      1025:1ee412f7fec9
user:        ysr
date:        Fri Oct 16 02:05:46 2009 -0700
summary:     6888898: CMS: ReduceInitialCardMarks unsafe in the presence of
cms precleaning

Bisecting the main hg repo shows it starts failing with 0491be4448bf.

----------
files: repro
messages: 11241
nosy: jcoomes
priority: bug
status: unread
title: incomplete bisection in 1.4 and later

____________________________________________________
Mercurial issue tracker <bugs at mercurial.selenic.com>
<http://mercurial.selenic.com/bts/issue1953>
____________________________________________________
-------------- next part --------------
A non-text attachment was scrubbed...
Name: repro
Type: application/octet-stream
Size: 490 bytes
Desc: not available
URL: <http://selenic.com/pipermail/mercurial-devel/attachments/20091215/de9057e9/attachment.obj>