Le Jeu 26 janvier 2006 13:50, Thorsten Leemhuis a écrit :
> Am Donnerstag, den 26.01.2006, 13:25 +0100 schrieb Nicolas Mailhot:
>> Le Jeu 26 janvier 2006 12:35, Thorsten Leemhuis a écrit :
>
>> My stupid way of doing things is to just create a separate srpm for all
>> branches (%{dist} is my friend), cvs-import them (whith the import
>> script
>> magically noticing already-uploaded sources and tagging everything as
>> needed) then make plague.
>
> There is one reason why I'm currently considering to forbid the use of
> cvs-import for already imported packages: Most people won't run a diff
> against the old package/the old spec this way before committing (or does
> the script force a diff before committing? I never used it for anything
> else than importing).
Actually my full workflow is cvs update package_I_need_to_change, modify
spec in devel, rpm -ba --define "dist .fc5" on result, import, sometimes
diff files with other branches
The big downside of doing check-ins manually is you're not pushing what
you've tested, and it's real easy to "forget" a file and get wrong
results.
With import of srpm generated in -ba you ensure not bit is forgotten.
(printing a diff of the result would be helpful so you can do a final
check before make plague)
> I run a "cvs diff -u" every time before I commit something and now and
> then (every fifth package round about) I find debug leftovers or small
> errors this way *before* committing them. Maybe I'm just to stupid and
> other people never make such errors -- but I suspect they do.
I'm stupid the other way - when doing fine-grained commits I always forget
stuff
--
Nicolas Mailhot
--
fedora-extras-list mailing list
fedora-extras-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-extras-list
