[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Gnu-arch-users] configs are too weakly coupled
From: |
Tom Lord |
Subject: |
Re: [Gnu-arch-users] configs are too weakly coupled |
Date: |
Fri, 25 Feb 2005 08:16:39 -0800 (PST) |
Acked for the bug database.
As a straw-man, how about:
% tla config-status CONFIGNAME
Report the existence of extra or missing required subtrees and
subtrees whose tree-version or patch-level is wrong.
% tla set-aside-extra-subtrees CONFIGNAME
Move all subtrees that don't fit CONFIGNAME to a ./+saved-trees
subdir of the top-level dir of the tree.
% tla restore-missing-subtrees CONFIGNAME
Fill the missing parts of a config in, taking trees from ./+saved-trees
if they are there.
Out of those, it should be possible to synthesize a very nice:
% tla switch-config FROM TO
which is:
% tla set-aside-extra-subtrees TO
% tla set-restore-missing-subtrees FROM
% tla in-each-config-subtree CONFIGNAME SUBCOMMAND [...]
As in:
% tla in-each-config-subtree CONFIG replay
with options to save output and error output in files (in each
subtree), stop at the first error or run in all trees in spite
of any errors, etc.
-t
From: "Robert Anderson" <address@hidden>
There is a recurring problem in my development group that stems
from the way that there is really no strong connection between
configs and the trees built from them.
It goes something like this:
Devguy: "I built after your merge and there's a link problem now."
Me: "Are you fully up-to-date? It builds for me."
Devguy: "Yes, everything is fully updated. No luck."
The problem is invariably that a config file has been changed,
and the devguy is unaware of this and is confused and frustrated
that he can't get up-to-date by issuing "update" commands.
Frankly, I think he's right to be. You shouldn't always have to
check out from scratch, or sift through commits wondering if
someone changed a config file, manually tweaking your trees if
there was, or write your own hacky scripts that verify the
consistency of configs and tree layouts. It ought to be a
one-off operation to verify that a config is fully checked in on
the one side, and a one-off operation to bring the config up to
date on the other, or verify that it is.
Maybe this is something the baz guys can think about.
Bob
_______________________________________________
Gnu-arch-users mailing list
address@hidden
http://lists.gnu.org/mailman/listinfo/gnu-arch-users
GNU arch home page:
http://savannah.gnu.org/projects/gnu-arch/