[Gnu-arch-users] Re: robin's per-directory regexps

[Tom Lord]

    >>>> What does tree-lint have to do with anything?

    >>> To trap .arch-inventories that are not tagged and source. [....]

    >> No, the patch makes them automatically "control files" so they are
    >> always source.

    > Backup a second here: You're altering the patch to not
    > synthesize a tag. So users will have to manually add a .id or a
    > tagline for the .arch-inventory to be committed to the archive.

If using `explicit'.   Otherwise a name tag is used.

And if explicit is used, untagged source is a hard (commit preventing)
error.  So there's no need for a special tree-lint case, is there?


    > If a) untagged-source precious is in use, and b) no tag has been
    > created, won't the .arch-inventory file in queston NOT get added to the
    > archive ? And won't it /still/ perform it's overrides during the
    > changeset creation?

    > Tla get on the result will then give you a tree that iimmediately fails
    > tree-lint.

That would be true except that, unless I'm overlooking something
stupid, you can't commit (in any tagging method) without including all
control files (because they automatically classify, by name, as
source).  In names-method trees, inline, and tagline -- you'll get a
names tag "for free".  With inline or tagline and --strict to commit
you'll get an error.  With explicit, you'll get an error even without
--strict.

-t