Re: [Gnu-arch-users] unmatched files

[Robert Collins]

On Mon, 2003-11-17 at 10:28, address@hidden wrote:

> I simply fail to follow you here, what on earth are you talking about..
> I find the issue Colin states very true, its very confusing to see tagging
> and regexpes fighting, certainly when its possible to _not_ commit explicitly
> added files because your regexpes say they are not elligable (see the
> relevant bugreport).
> 
> I would appreciate an answer to colin's question that would make me see
> your way, so maybe I don't have to curse soo badly the next time this bites
> me.

There is a bug in tla add today, in that you can tla add <foo> where foo
doesn't pass the source regex. 

Secondly, when you change regex's, inconsistencies can occur: but tla
tree-lint, which commit calls will detect that.

In terms of concept, the logic as I understand it:

By classifying the contents of a tree into various categories, we can
perform automated operations on them: errors on files that should never
exist, automatic purging of backups and (truely) temporary data.

tla has a distinct sub layer that performs this classification. It's
part of inventory, but not the totality. The tagging of files via tla
add / embedded tags depends on this layer: it only tags and examines
source files and directories.

Now, the reason that regexs trump tags is simple: the tagging layers on
top of the source finding code.



Rob
-- 
GPG key available at: <http://www.robertcollins.net/keys.txt>.

signature.asc
Description: This is a digitally signed message part