mergepoint and null delta results (was Re: [cvsnt] Latest Updates)
news at derived-software.ltd.uk
Thu Feb 5 18:44:30 GMT 2004
On 2004-02-02, Phil Richards <news at derived-software.ltd.uk> wrote:
> On 2004-02-02, Harrison, Andrew <andy.harrison at anite.com> wrote:
> > Cool, I guess that was prompted by my questions last week? There was some
> > concern then that this patch "might break things". Can I ask what things
> > might get broken and how they would break?
> Nothing *should* be broken, but just keep an eye on merges back and forth
> between branch and mainline. It works in all the cases I've tested, but
> there are possibly edge cases I haven't tested (deleted, resurrected file
> in various combinations).
Well, I've done my first "production" two-way merge and had mostly
positive results. The merging (as before) worked fine, and no null
deltas were made.
The only glitch relates to deleted files.
For the file in question:
1. A branch was made (call it B1).
2. The file was updated on the mainline (ML).
3. ML --> B1 merge performed, and committed on B1.
4. The file was deleted on ML.
5. ML --> B1 merge performed.
At (5.) a warning message is printed stating that the file has been
deleted on the ML but modified on the branch.
This is true - but one might hope that the server would notice that
the only modification on the branch was a merge from the mainline,
and that the version on the branch was the same as the deleted one
on the mainline, and mark it as "removed".
There shouldn't be a problem as long as you take notice of the warning
messages from CVS. If you don't, you are probably making a mess of
things already :-)
p.s. The new (beta) WinCVS has some cool mergepoint arrows in the graph
view now. Neat!
change name before "@" to "phil" for email
More information about the cvsnt