Re: [Gnu-arch-users] Re: arch on windows?

[Harald Meland]

[John Meinel]

> Adam Sampson wrote:
> | John Meinel <address@hidden> writes:
> |
> |>So if all we do is switch from {arch}/c/c--b/c--b--v/c--b--v--p-nn.* to
> |>{arch}/c/--b/--v/--p-nn.* (I'm leaving in the -- because 'b' can be
> |>empty) we'll probably have a 75% solution.
> |
> | I think I'd prefer {arch}/c--/b--/v--/p-nn.* -- filenames starting
> | with "--" have the potential to irritate people even more than those
> | starting with "{".
>
> Fine by me. I agree, -- can get you into trouble, thanks for pointing it
> out.

The c--/b--/v--/p-N scheme could still cause trouble for "null branch"
versions; for instance, bash interprets a bare "cd --" as "cd $HOME".

Also notice that a single trailing dash should suffice (although bash
interprets a bare "cd -" as "exchange $PWD and $OLDPWD"), and that the
"p-N" part needs to distinguish between (at least) "base-0",
"patch-N", "version-N" and "versionfix-N".

(About the "at least" part: Can we safely assume that arch won't add
even more p-N variants?)

I'm not sure how much effort should be put into avoiding the
occasional shell trouble with null branch versions, but if we can find
some not-allowed-at-the-end-of-{category,branch,version} character
that also doesn't cause either shell or arch trouble when appearing at
the start or end of filenames, we could use that instead.

Off the top of my head, I can't think of any such characters.

An alternativ schem, that hopefully would avoid any shell trouble,
could be

  C$category/B$branch/V$version/R$revision.*

  Example: Ctla/Bdevo/V1.3/Rpatch-28.*

An additional bonus with this scheme is that it would preserve the
"take one look at a directory name and you know what depth you're at"
property of the current scheme; in fact, it would even improve on the
current scheme, as it would distinguish the category and branch levels
when the branch name is empty.

However, I suspect some people might say it's too cryptic, or
inconvenient for shell scripting.  If that's a problem, we could
consider striking a better balance between the prefix length in each
component and corresponding readability/parseability -- e.g.

  ctg-$category/bra-$branch/ver-$version/rev-$revision.*

Cheers,
-- 
Harald