Re: [Gnu-arch-users] New aba command: revert, name clash with tlash

[David Allouche]

On Fri, Mar 26, 2004 at 03:19:33PM -0500, Aaron Bentley wrote:
> Actually, for "reject", I'd rather have a log-preserving replay 
> --reverse, as in
> 
> $ aba star-merge ^foo
> $ aba reject ^foo--patch-3
> $ aba reject ^foo--patch-5
> $ aba commit -s "merge from foo (dropped patch-3, patch-5)"
[...]
> 
> While "revert" and "reverse" are cognates, the verb form of "reverse" is 
> actually "reverse".  "Revert" means "return to an earlier state", as in 
> "revert to saved version".  Since "replay --reverse" can retain some 
> changes while removing others, I think "reverse" would be a better name 
> for the "replay --reverse" alias.

You are completely right. I modified tlash to use "reverse" instead of
"revert" for "tla replay --reverse".

> >The ability to use and view only the patchlog or non-patchlog part of
> >changesets is a generally useful feature which I believe is okay with
> >Tom to put in tla.
> 
> It makes sense to me too.  We can hack around it for specific cases, but 
> it seems tough to generalize.

I do not think that is hard to generalize, assuming non-pathological
changesets, and pathological changesets can be easily detected. But that
would be a different discussion.

> I'd also prefer to ignore changes to ++default-version.

Is it actually possible to modify this file in changesets???  If yes,
would tla actually produce such changesets w/o having to hit it
repeatedly on the head?

In any case, I think such a changeset would definitely fall in the
pathological category...

-- 
                                                            -- ddaa