Re: [Gnu-arch-users] [MERGE REQUEST] unify various commands around arch_fqvsn_from_tree_and_input

[Aaron Bentley]

Colin Walters wrote:
> Interesting.  Well I think we're actually solving somewhat different but
> related problems.  Notice that we actually patched different sets of
> commands (you patched tag and sync-tree, I patched delta, get-changeset,
> and undo).

I think get-changeset is an arch_determine_revision candidate, not a 
fqvsn_from_tree_and_input candidate, because it's not a project tree 
command, and not necessarily run in a project tree.  If you look at get, 
you'll find it uses the archive-latest revision, and I think 
get-changeset would be analogous.

It would also be nice to stick an arch_check_for in there, to handle the 
case where the user-supplied revision/version doesn't exist.

> Let me describe my function arch_fqvsn_from_tree_and_input.  It's meant
> to take any sort of user input (from a fully-qualified revision like
> address@hidden/project--main--0--patch-32) to bare patchlevels like
> patch-32.  If the input is already fully-qualified, just return that. 

Err, are you handling the case where you just get a package-version and 
you have to determine the revision from the logs?  (see 'tla changes' 
for an example)

> The key difference with what arch_fqvsn_from_tree_and_input does is that
> it uses the current tree version to disambiguate.

Okay, so little risk of us making conflicting changes.  (Though I've 
patched delta too...)

> Now I'm pretty sure I'm on the right track with this thing because most
> of my changesets consist of *deleting* code and replacing it with this,
> with no loss in functionality or semantics, etc.  

Just be careful not to change functions that already use the 
archive-latest revision!

> With the sole
> exception of delta, and I'm about to fix that now.

Perhaps a soft_errors parameter would do the trick.


> Your function doesn't touch the tree version at all - instead it looks
> at the default archive.  So I think both of them make sense.  Probably
> :)

Yeah, my function was basically a rip of sync-tree's functionality.  But 
it only uses the default archive if there's no archive name provided by 
the user.

Aaron