FUNKY tags.

I just tried a cvs->git conversion using the git-cvsimport-script
and cvsps flagged a bunch of tags as **FUNKY**

I've no idea what I did when I tagged those trees, but according
to a google search, cvsps does that when it find patchsets which
are chronologically (and thus by patchset id) earlier than the tag,
but are tagwise after.  Spooky.

I looked the tree over with gitk, and it seemed fine around those tags,
so I'm wondering what value there is in preserving this info, and
whether we should remove it at cvsimport time somehow. Comments?

		Dave

On 8/17/05, Dave Jones [off-list ref] wrote:

I've no idea what I did when I tagged those trees, but according
to a google search, cvsps does that when it find patchsets which
are chronologically (and thus by patchset id) earlier than the tag,
but are tagwise after.  Spooky.

It's probably tags that were moved around with "cvs tag -F footag".
When using cvs in with a dovetail strategy, people tend to merge
BRANCH->HEAD and use a floating tag to mark how far it's been merged
in.

I am somewhat worried about cvsps getting confused by these floating
tags. Any help in teaching cvsps to ignore tags is welcome ;)


m

On Wed, Aug 17, 2005 at 12:55:18PM +1200, Martin Langhoff wrote:
 > On 8/17/05, Dave Jones [off-list ref] wrote:
 > > I've no idea what I did when I tagged those trees, but according
 > > to a google search, cvsps does that when it find patchsets which
 > > are chronologically (and thus by patchset id) earlier than the tag,
 > > but are tagwise after.  Spooky.
 > 
 > It's probably tags that were moved around with "cvs tag -F footag".
 > When using cvs in with a dovetail strategy, people tend to merge
 > BRANCH->HEAD and use a floating tag to mark how far it's been merged
 > in.

In my case, at least the most recent of those cvs tag operations
was just a 'cvs tag x86info-1_14'. Nothing fancy.  I'm fairly sure
there was nothing fancy about the earlier instance either.
So sure in fact, I had to look up that -F flag in the man page to find
out what it did.

x86info cvs shouldn't even have any branches, so theres no strange
head merging going on.

 > I am somewhat worried about cvsps getting confused by these floating
 > tags. Any help in teaching cvsps to ignore tags is welcome ;)

If you're curious, its a sourceforge project at cvs.sourceforge.net:/cvsroot/x86info
module is 'x86info'.  Both x86info-1_14 and x86info-1_6 are marked as funky
for reasons I'm unaware.

		Dave

On 8/17/05, Dave Jones [off-list ref] wrote:

In my case, at least the most recent of those cvs tag operations
was just a 'cvs tag x86info-1_14'. Nothing fancy.  I'm fairly sure
there was nothing fancy about the earlier instance either.
So sure in fact, I had to look up that -F flag in the man page to find
out what it did.

Ok - that reduces the suspects to 
 - a tag applied to only one part of the tree
 - a tag that took longer to apply than fuzztime (large tree, slow cvs
connection?)
 
let me know, and if I find time I may look at the repo. 

martin