monotone - Bugs: bug #13178, cvs_import still produces wrong...
You are not allowed to post comments on this tracker with your current authentication level.
bug #13178: cvs_import still produces wrong data on branch (continuation from 13032 and 13063)
Submitter: | None | ||
Submitted: | Tue 24 May 2005 09:20:52 PM UTC | ||
Category: | storage system | Severity: | 3 - Normal |
Item Group: | incorrect behavior | Status: | None |
Privacy: | Public | Assigned to: | None |
Open/Closed: | Open | ||
mtn version --full: | monotone 0.19 (base revision: 0dad47afa920a307999d13eda117836b15cd4d82)
|
Sun 12 Jun 2005 08:04:04 PM UTC, comment #2: |
graydon hoare <graydon> |
Sun 12 Jun 2005 07:07:39 PM UTC, comment #1: I have just downloaded a fresh copy of monotone, and it seems like this problem has been fixed. Here is the version information for what I'm running
|
Stephen Peters <smpeters> |
Tue 24 May 2005 09:20:52 PM UTC, original submission:
Since anonymous users cannot add comments, I've to create yet another bug.
|
Anonymous |
Depends on the following items: None found
Items that depend on this one: None found
There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.
Follows 1 latest change.
Date | Changed by | Updated Field | Previous Value | => | Replaced by |
---|---|---|---|---|---|
2005-05-24 | None | Attached File | - | Added e.tar.bz2, #2540 |
Powered by Savane 3.13-f8d8.
Corresponding source code
the precise bug in question has been solved, but there is actually a much more subtle follow-on bug which has been identified here.
essentially the root of the remaining problem (which is not reported in this BTS but which I've received in private email) is that monotone's CVS importer uses "best current" knowledge about a commit, while processing each file, to infer the global commit points. this means that if monotone processes files A and B, in order, and B contains information which would help disambiguate commits in A, the information is not used, because it is already finished consulting file A. ultimately this leads to lower-fidelity imports than we would like.
the solution is simply to collect the information in a non-condensed form first, analyze it globally across the entire repository, and then condense it into individual commits. this is, note, how cvs2svn works (I've been studying it today); but it requires quite a bit of rewriting, so it will be a couple weeks before I can confidently say positive things about this bug.
please leave this bug to me. I will stay on it until it is fixed.