Re: [Gnu-arch-users] Moving a sourceforge project to arch

[Michael McCracken]

Thanks everyone for the advice.

I was talking about people sending me tarballs, as well as people 
saying "I would contribute, but I don't want to learn CVS". I can't say 
I like it, but I'd like to accommodate them. So they don't want to use 
cvs diff, but they might be happy with a script like 'send_changes' or 
something. Essentially, I have non-programmers wanting to try their 
hand at hacking on my code, and I want to make it easy, but I don't 
want to have to swim through tarballs. That said, Charles' suggestion 
should be useful.

It sounds like the choice for my project would be to write a script for 
getting the source and another for sending changes, then I can migrate 
those scripts to arch without contributors needing to relearn stuff. I 
imagine I can probably use #1 from Brian and if people want to add or 
rename files, they can learn more about tla.

Actually, #2 sounds better, but I didn't understand what was bad about 
being binary.

Also, thanks Miles for the info about tla-cvs-sync. I will look into 
that, it sounds great.

-mike

On Feb 26, 2004, at 4:16 PM, Charles Duffy wrote:

> Presuming you're talking about people sending you tarballs (the one 
> case
> that Brian May didn't really handle):
>
> Arch can handle this kind of thing nicely, especially in tagline mode
> (where you don't need to worry about them doing things like giving you 
> a
> tarball with renamed files but untouched tags). You can just [for
> instance] clear out your old source files from a copy of your tree (tla
> inventory -s | xargs rm), unpack the tarball they sent on top of it, 
> and
> use the regular commands to see what they did (tla changes), maybe use
> your regular tools to back out bits you don't like (vimdiff $FILE `tla
> file-find $FILE`), and so forth.
>
> There's probably better idioms for these things someone else can
> suggest, too.


On Feb 26, 2004, at 2:58 PM, Brian May wrote:

> I get the impression from the above that you mean that they will send
> you the entire tree, not just the patches?
>
> Why do they not want to do commits? Is this because it is too hard, or
> is it because they are scarred of making mistakes and/or making
> changes you dislike?
>
> If contributors don't want to use cvs diff to create patch files, I
> don't think arch can help. It means you would have to isolate what
> changes they made, and propagate them across to your tree.
>
> If contributors do use cvs diff though to generate patches, there seem
> to be two ways they can do the same thing with tla:
>
> 1. tla changes --diffs | mail -s subject ...
>    (could be done via script)
>
> 2. tla changes -o tmpdir && tar -czvf tmpdir.tar.gz tmpdir
>    send tmpdir.tar.gz via MIME mail.
>    (could be done via script)
>
> 3. distribute scripts that will automatically create a local archive
>    and branch your tree to the local archive, and mirror it to a
>    publicly accessible location.
>
> 4. tla commit to central location. Obviously this may not be a valid
>    option here.
>
> (not tested)
>
> I believe 1. is easiest to do, but I think you will have problems with
> renamed/moved files, 2 will work with renamed and moved files, but the
> result is binary. 3 seems to be the preferred approach, however, it
> requires each contributor have a publicly accessible mirror.

smime.p7s
Description: S/MIME cryptographic signature