[issue1531] bogus convert-repe entries in hg convert messes up diffs in hg serve

Brad Penoff mercurial-bugs at selenic.com
Tue Mar 3 14:09:48 CST 2009


New submission from Brad Penoff <penoff at cs.ubc.ca>:

I am seeing a problem with hg covert (v 1.1.2).

I'm working on a project with some others and they manage the CVS repository.  I
have a script that rsync's this repo, does a local CVS checkout, then uses "hg
convert".  I run this every 15 mins and have an "hg serve" running where I can
nicely see the patches. This same set of scripts has worked for me on several
projects that I'm involved with.  But not this most recent one....

Now, when someone upstream has committed something, when "hg convert" runs to
get only the new patches, it inserts these seemingly bogus convert-repo entries
between each patch. 

If I destroy my entire local hg repo, and re-run the script from scratch, those
convert-repo entries are magically gone.  But obviously that is non-ideal and it
takes more time (basically eliminates the point of the rsync).

I was wondering if anyone had ever seen this bogus convert-repo entry phenomenon
when using hg convert from CVS with datesort and if there was a way around it.

The funny thing is, if i delete everything and run it all from scratch, those
entries are not there.  It's not just the bogus entries that are the problem...
viewing the diffs becomes faulty; it just shows its diff between the file and
/dev/null (so not so useful).  I don't really care about the entries per se... 
it's the inability to view the diffs, because that's what i'm doing all this
converting for in the first place.

I could be doing something wrong but i do the same exact steps for about 6 other
projects and have never seen this.

The bogus convert-repo entry is always and only 'default'.  Email me if you
would like to see an example (I can't pass the link now because I always am
starting over from scratch so that I can view the diffs, so to ensure there is a
bogus entry, I'd have to avoid doing this).

Here's an example bogus entry, copied from hg serve entry...

1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000  
1.2 +++ b/.hgtags Tue Mar 03 18:36:07 2009 +0000  1.3 @@ -0,0 +1,1 @@  1.4
+f47a9d80b4de6f79b83151b181a778709eeb2e66 default

----------
messages: 8721
nosy: penoff
priority: bug
status: unread
title: bogus convert-repe entries in hg convert messes up diffs in hg serve
topic: convert

____________________________________________________
Mercurial issue tracker <mercurial-bugs at selenic.com>
<http://www.selenic.com/mercurial/bts/issue1531>
____________________________________________________



More information about the Mercurial-devel mailing list