UNIX for Dummies Questions & Answers

I'm in a fustrating situation where I am repeatidly checking code, editing,
synchronizing, finding something is broke, reverting all my changes and starting over. This if often easier than trying to merge my changes with someone who has beat me to the checkin.

Is there a way I can mitigate this with a script? Doing perl search and replace on files helps. I know there are some patch utilties that will generate a patch file that (I think) are part of CVS. The problem is that we are not using CVS or SVN.

Is there an open source utiltity, perhaps a perl module, that will apply the changes I have made to a file again when I have starting over from a fresh checkout? I was hoping for one based on search strings instead of line numbers.

It seems like there should be a way to make emacs save a journal that I can execute again later.

Thanks,
siegfried

If it's not SVN or CVS then what are you using?

Sounds like you should look at some of the distributed version control systems out there. Some have pretty sophisticated patch handling. Darcs is supposed to be good at this, although I can't say I've had it live up to its promise for me. (I like it a lot, otherwise.)

I am told we are using a derivative or variant of perforce called source depot. I don't know how this could be because I've used perforce before and this command line interface is quite different. Also, I don't understand how a derivative or variant of a commercial product like perforce could exist.

I do have rcs installed for my personal use: that must be able to compute and execute deltas.

Anyway, I did not select source depot. I'm just "contingency staff". It's command line interface sure is crude compared to the nice GUIs available for CVS and SVN.

My prediction came true again today! Only this time, it is because I broke the build and I have to revert everything to fix that immediately.

So I did my editing with using elisp code and saved that code so I could execute my editing changes again. There must be a better way!

Siegfried

You certainly can use RCS locally if Source Depot will allow that. You don't even have to have the ,v files in the same directory or a subdirectory, although I'm not sure this is documented anywhere -- I saw a colleague use that and vaguely recall having read about it somewhere, but I could not find it in the documentation when I wanted to read about it.

(This is from memory, I don't have RCS here.)

Code:

co file /path/to/file,v

Similarly for ci, rlog, etc.

If the conflicts are complex, you still end up with a mess. I hear Mercurial (or was it Monotone?) is supposed to bring some relief to resolving conflicts. Also, have you tried xxdiff, emerge (if you use Emacs) or some other dedicated diff merge tool?

Of course, according to theory, you should talk to your colleagues so you can avoid conflicts, not just struggle with a merge tool in your cubicle.

Can you help me understand the rcs commands to compute the deltas between two files and save the delta so I can apply it again after I have reverted using source depot?
Siegfried

It's all just diff and patch anyway, RCS doesn't particularly add anything to that. I have been assuming you've tried diff and patch already and not been happy with the result? If not, try that first.

Code:

mv file file.mine
sd revert file # or whatever the command to revert is
diff -u file file.mine >mine.patch
sd get # or whatever you do to bring in the latest version
patch <p0 mine.patch
# pore over conflicts
sd commit
rm file.mine mine.patch

It's been a long time since I used RCS, but if I remember correctly, even CVS brings substantial improvements to conflict handling in particular. There's nothing to prevent you from checking the files into a local CVS repo, save for the CVS directory which might upset another version control system, and/or need manual sync if you use another version control in tandem.