Date: 12 Jul 1995 12:08:57 +0800 From: peter@haywire.dialix.com (Peter Wemm) To: freebsd-hackers@freebsd.org Subject: Re: CVS 1.5 is out Message-ID: <3tvhsp$amq$1@haywire.DIALix.COM> References: <Pine.BSF.3.91.950711165211.7169C-100000@minnow.render.com>, <199507112142.PAA06239@rocky.sri.MT.net>
next in thread | previous in thread | raw e-mail | index | archive | help
nate@sneezy.sri.com (Nate Williams) writes: >> I just noticed that CVS 1.5 has been released and one of the things it >> mentions that is new is client-server support. Does anyone know whether >> this is any good? >The client-server stuff is 'OK'. I've been testing it for a while now, >and I even set it up with a machine in a the Bay Area and did some >commits to/from Montana to see how things work. Basically, my *opinion* >is that the sup/CTM synchronization with a local CVS repository works >better for development than the client-server stuff in CVS 1.5. Yep. It works.. It isn't perfect (we've been using it for a while now between our Sydney and Perth sites (on opposite sides of Australia for the .au challenged.. :-). I'm quite satisfied with it - it's just survived some serious beating-up while doing some merging of about six different branches of INN. One very big thing in it's favour, is that it when updating files, it will send a diff instead of the new version from the repository host, and a md5 of the result. Only if the patch fails, will it send the while file. >Basically, the biggest downside of the current setup is that your >network link *must* be up for *every single* CVS operation, which makes >things obnoxious for doing development over slow SLIP/PPP links over the >internet. This is true, although using diffs helps this somewhat. >However, I am using it for keeping a bunch of machines on my local >network synchronized, and it works very well for that. >> In particular, is it good enough to reliably support remote commits >> across the internet without screwing up big time when the link goes >> down mid-commit? >It's pretty safe now. Basically, they do things in a 'transaction' >style which means that all of the information is completely passed to >the server before anything is done. This avoids problems with links >going down or partial commands locking up the tree. There was a patch submitted about a day before 1.5 was released that closes off the last remaing place where locks could be left behind, but it was too late in the release cycle to make it for 1.5 - I dont remember seeing if it was applied in the current snapshot. >Nate For what it's worth, I'd personally prefer to use a cvs-1.5 type method of access than a daily sup. Cheers, -Peter
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3tvhsp$amq$1>