[issue276] common ancestor bug

[Hollis Blanchard]

New submission from Hollis Blanchard <hollis at penguinppc.org>:

In Xen merges, I've frequently been getting messages like these:
 local changed tools/vtpm/tpm_emulator-0.2b-x86_64.patch which remote deleted
(k)eep or (d)elete?

hg log tools/vtpm/tpm_emulator-0.2b-x86_64.patch:
changeset:   7093:06d84bf87159
parent:      7092:97dbd9524a7e
parent:      7070:2f83ff9f6bd2
user:        djm at kirby.fc.hp.com
date:        Thu Sep 22 11:42:01 2005 -0600
summary:     Merge latest xen-unstable into xen-ia64-unstable

changeset:   7019:55fc0ecc19c3
parent:      7018:9647be59212d
parent:      7001:7a45b8ccef01
user:        emellor at ewan
date:        Wed Sep 21 15:25:58 2005 +0100
summary:     Merge.

changeset:   6998:492fe12a6db4
user:        kaf24 at firebug.cl.cam.ac.uk
date:        Wed Sep 21 09:56:34 2005 +0000
summary:     Fix .hgignore and add missing vtpm patch file.

All of these are upstream changes; i.e. the file has not in fact been modified
locally, contrary to the prompt. Digging in to localrepo::update() (in hg
version 0.9), I discovered that the common ancestor calculation seems to be
incorrect.

The attached script illustrates the problem. After it completes, in the clone
directory, "hg debugancestor .hg/00changelog.i 5 4" should yield 3, the "merge"
changeset. It incorrectly yields 2.

----------
files: ancestor.sh
messages: 1443
nosy: hollisb
priority: urgent
status: unread
title: common ancestor bug

____________________________________________________
Mercurial issue tracker <mercurial-bugs at selenic.com>
<http://www.selenic.com/mercurial/bts/issue276>
____________________________________________________
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ancestor.sh
Type: text/x-sh
Size: 413 bytes
Desc: not available
Url : http://www.selenic.com/pipermail/mercurial/attachments/20060531/0bae403a/ancestor.sh