Re: anoncvs vs cvsup

[naddy@mips.rhein-neckar.de (Christian Weisgerber)]

Ron 'The InSaNe One' Rosson <insane@lunatic.oneinsane.net> wrote:

> My past posting to the misc mailing list were about having a problem 
> with doing a make build after I had created a "src tree" using cvsup. 
> Well I reverted back to try anoncvs and did a make build and it went 
> smoothly.

The last time I tried this, the problem was that cvsup didn't
preserve the permissions of files while cvs did, and the build
process depended on some files being executable. Since then some
fixes to that respect have gone into the tree, but I don't know
whether they are complete and how many of those are already in 2.6
as released.

Just why the update by cvsup loses the permissions is a mystery to
me. cvsup itself is very much capable of exactly mirroring file
permissions. How does cvs store file permissions? Does it go by
the permissions of the repository files or are the permissions
stored as metadata somewhere?

cvsup.de.openbsd.org I'm cvsupping from updates its repository by
ctm. But from looking at a few ctm deltas I happen to have lying
around, ctm clearly transmits the file permissions as well.

> The only problem I have is that the cvsup takes about 30-45
> minutes on my little ole 56K Frame Relay link and anoncvs takes about 
> 12 hours.

I don't know whether anoncvs bottlenecks due to server speed, client
speed, network speed, or protocol misdesign, but it certainly is
significantly slower than cvsup. On my FreeBSD boxes, I use cvsup
for *local* updates from the repository because of the superior
speed. (Well, you want to have the checked out tree and the repository
on separate spindles.)

-- 
Christian "naddy" Weisgerber                  naddy@mips.rhein-neckar.de