Re: [gentoo-portage-dev] [PATCH] elog-base

[Thomas de Grenier de Latour]

On Sun, 23 Oct 2005 11:54:07 +0900
Jason Stubbs <jstubbs@g.o> wrote:

> Seeing that interspersed einfo and ewarn calls will be separated
> by class anyway, is there any need to play with the ordering at
> all?

That's also one thing i don't understand with elog: why does it
separate messages in many files instead of using a single one with
a simple markup? In the implementation of a similar feature i had
done as part of bug #37491, i was using something like:

...
[ebuild:/path/to/foo-bar-1.2.ebuild package:cat/foo-bar-1.2
step:post_inst date:1234567890]
[emessage:info]
You can try to run "foo-bar --break-things", it's fun.
[emessage:warn]
Don't do it too often tho, it hurts.
[emessage:info]
Also, run "foo-bar --check-things" from time to time.
...

It was really easy to parse and pretty-print, and was giving
messages in the right order. And in some cases, it matters: for
instance, the following would have been missleading: 
 * You can try to run "foo-bar --destroy-things", it's fun.
 * Also, run "foo-bar --check-things" from time to time.
 * Don't do it too often tho, it hurts.


Also, one thing i had seen when working on this old patches was that
einfos from eclasses really clutter the logs you get. I mean, few
packages don't have at least an "Applying something.patch" or a
"Cleaning orphan /usr/lib/python/plopplop.pyo", etc.  I think it
would be a good idea to provide some einfo-like functions which do
not write to logs, just for eclass writers who want color when they
feel the need to tell the world what their code is currently doing.
That would reduce a lot the number of packages with false-positive
elogs. 

-- 
TGL.
-- 
gentoo-portage-dev@g.o mailing list