[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Gnu-arch-users] the behavior of tla get
From: |
Andy Tai |
Subject: |
Re: [Gnu-arch-users] the behavior of tla get |
Date: |
Tue, 22 Nov 2005 06:50:32 -0800 (PST) |
Derek, please go ahead and post the patch... As for the penalty parameter, a
reasonable default
would be good. I think many users do not need to adjust that value--people
would complain again
how hard it is to optimize the behavior.
--- Derek Zhou <address@hidden> wrote:
> Hi all,
> Right now "tla get" can run into two problems:
> 1, if there is no revlib, tla will get the nearest cacherev and patch
> on; ignoring the possibility of having some earlier local version;
> 2, if there is a revlib, but does not contain any useful info, tla will
> go all the way back to the import or the first tag into this archive and
> patch on. Ignoring the existance of some cacherev in the middle.
> Neither one is very great. It would be better if there is a way to trade
> off between local cache (revlib) or remote cache (cacherev), by some
> user customizable parameter.
> I have made a rather simple heuristic:
> A patch has a cost of 1; a full cacherev has a cost of eg. 50. The
> client can search backward to find out the minimum cost to get a
> revision.
> I get it mostly working, except for the case of greedy but not sparse
> revlib. Any comments?
> Right now I have my parameter hard coded. Obviously we want this number
> customizable. Where should we put it? Maybe a file like
> {arch}/=cacherev_penalty or something?
> Derek
>
Andy Tai, address@hidden
Free Software: the software by the people, of the people and for the people!
Develop! Share! Enhance! Enjoy!