Jason, since you've done this before make you can help me understand to what degree we should strive
to keep these files in sync. I notice that many of the differences between Trunk and the new 1.2
Branch are related to omissions of $Rev and $Date in various java, js, jsp, and properties files.
Are these entries being added automatically by either SVN or an IDE, and should we bother syncing
files with only these differences ?? Thanks
Tim
Jason Dillon wrote:
Yes, this is the best way... merge from 1.2 to trunk, as *most* of those
changes will be fairly simple to apply, and automatic with SVK (well, up
until the point when we rearrange trunk, but until then).
But some minor changes may also need to go the other way. SVK should be
able to handle this. When I was working with SVK for the m2 migration
branch, I was keeping all svn notifications I got, then when they
buffered up enough, I would use SVK to merge each change, limiting the
path to either file or src/main/java for the modules affected to avoid
pulling in unwanted changes. In the case of the m2 migration, unwanted
changes would be stuff in a pom. You could do a merge from the branch
root, then cherry pick the changes, but that is not much fun when there
are a bunch of changes.
Anyways... IMO its best to try to only merge from 1.2 to trunk, and if
needed only merge from trunk to 1.2 on a per-file basis.
That means if you are working on fixing a bug, its best to fix it in
1.2. But experience has shown that people will work off of trunk and
merge into branches more often than desired. But, if you are careful
about the merge then no major problems should pop up.
I also recommend, when using svk smerge, that you first run with -C to
see what it wants to do first. Limit the changes pulled in to one merge
if possible to avoid picking up something you did not want.
And when you do the merge, use the -I flag to include the original text
of the commit into the merge automatically, this makes it easier to
track... and more automated... as if there are not conflicts, the merge
will not require any user interaction.
When you initially setup the svk config you will need to use --baseless
on the first smerge, but only for the first... all afterwards SVK should
have enough details to find the base, not sure what will happen if you
keep using --baseless, so I don't recommend it.
And if you run into anything strange, unlikely but might happen, hope in
#svk on freenode and ask, they have been very helpful to me.
--jason
On Dec 4, 2006, at 4:03 PM, Tim McConnell wrote:
Ok, I'm setting up SVK so that we can keep changes between the new 1.2
Branch and Trunk in sync. I don't mean to be too simplistic but I
would like to verify these assumptions on my part are correct (before
I do anything untoward):
1. The primary intent will be to ensure that changes made in the 1.2
Branch will get merged into Trunk. Ideally these will be fixes and/or
enhancements that have been made to the 1.2 Branch that must also be
ported into Trunk as well.
2. Changes made to Trunk will NOT be merged into the 1.2 Branch. This
should pretty much be business as usual (it would be very difficult to
try to identify just code fixes in Trunk that have to be retrofitted
back into the 1.2 Branch).
This seem reasonable to everyone ??
Thanks much
Tim
