Re: [Gnu-arch-users] Re: Default version for star-merge

[Tom Lord]

    > From: address@hidden (James Blackwell)

    > [*BIG* lightbulb]: 

    > tla star-merge (regex)

You keep coming up with applications for that "permit a regexp"
pattern and I keep liking them.   

    > and only perform the star-merge if the regex matches exactly one line in
    > =merged-from (or some other file that lists fully qualified versions
    > we've merged from)

And, heck, that's also a place where we can give "logical role"
nicknames to the related branches, allowing the command set to reflect
local policies and practices.   E.g.: 


        % tla star-merge /upstream/

        % tla star-merge /ssa-work/

        etc.

where the mapping from names like "ssa-work" to particular branches
can vary over time and possibly diverge from the branch labels.


    > Oh. I like that idea. It works both ways and I could *easily* imagine 
    > using that for a variety of archive names. 
    > 
    > Even better. Imagine *this* :

    > tla star-merge jblack/t.*rbrowse
    >  ( replacing tla star-merge address@hidden/tla--rbrowse )

    > tla get lo.*4/tla--d
    >  ( replacing tla get address@hidden/tla--devo )

    > tla rbrowse lord.*/tla
    >  ( replacing tla rbrowse address@hidden/tla )

    > Whadaya say?

I say: "thank you".



    > Tom Lord again:
    > > For another thing: the pointer to the other branch might not be all
    > > that we want to record.   For example, there may be a rule for
    > > how/where to submit a merge request back to upstream.   Perhaps these
    > > additional but related parameters should be planned for here as well.

    > Are you worried that might make arch a little too smart for its own
    > good?

Hopefully not.   I think ultimately it just automates a common process
that's otherwise useful to do "by hand".   That's pretty much all arch
does in the first place :-)

-t