From: Jeff D. <jd...@ka...> - 2001-07-09 22:09:06
|
As many of you know, Bill Stearns has generously dedicated a machine (known variously as {uml,user-mode-linux,uml-pub}.ists.dartmouth.edu) at Dartmouth ISTS to support UML development. Sourceforge is flaky enough that I'm planning on moving my primary base of operations from there to Bill's box and using SF and Rik's machine (ftp.nl.linux.org) as secondaries. Downloads are available from all three locations now. My next step is CVS. I think I have a recipe for using the ISTS box as a primary CVS repository and the others as mirrors: locally check out the ISTS UML tree check out the SF tree rsync from the ISTS tree to the SF tree check the changes back SF with something like cvs ci -m "`date`: daily update" This would keep everything up to date, but you'd lose the real log messages and tags from the SF CVS tree. If ftp.nl.linux.org is running CVS and I can rsync to it, then all that will be preserved there. SF, afaik, only allows CVS updates to CVS. I think that a sufficiently unholy script could be written to figure out what files changed, what the log messages and tags for each one are, and check them in so that all that stuff is preserved and if anyone wants to contribute it, I'll have a look at it. So, the upshot is that the ISTS CVS repository will be the official one with real logs and tags, the SF repository will be a mirror with no real logs or tags, and ftp.nl.linux.org could be either or none, depending on what services Rik has running on it. My main question is whether anyone has any problems with this scheme? Also, I'd like to know if I'm doing anything particularly boneheaded here, since I'm far from a CVS guru. Jeff |
From: Henrik N. <hn...@ma...> - 2001-07-10 11:45:11
|
Jeff Dike wrote: > My next step is CVS. I think I have a recipe for using the ISTS box as a > primary CVS repository and the others as mirrors: > locally check out the ISTS UML tree > check out the SF tree > rsync from the ISTS tree to the SF tree > check the changes back SF with something like cvs ci -m "`date`: daily update" It is more to it than that... but I have a script for mirroring one CVS tree to another CVS server... <http://squid.sourceforge.net/devel.sync>. Use it to mirror cvs.squid-cache.org to cvs.sourceforge.net/cvsroot/squid/ There is also a number of other scripts for the purpose of branched development with "floating branches" (the branch stays current with it's "mother"). See http://squid.sourceforge.net/CVS.html > This would keep everything up to date, but you'd lose the real log messages > and tags from the SF CVS tree. Same here. But not a problem for us (Squid) as the SF repository is another repository with a different goal. In our case the SF repository is for development and experiments relative to the main tree. Tags in the main tree are not relevant to the SF tree. > If ftp.nl.linux.org is running CVS and I can > rsync to it, then all that will be preserved there. SF, afaik, only allows > CVS updates to CVS. True. SF have promised to deliver ssh access for project managers to the CVS server for ages.. > I think that a sufficiently unholy script could be written to figure out what > files changed, what the log messages and tags for each one are, and check them > in so that all that stuff is preserved and if anyone wants to contribute it, > I'll have a look at it. Already done ;-) > My main question is whether anyone has any problems with this scheme? Fine by me. You could also consider doing what we do on squid.sourceforge.net, giving just about anyone who like to hack on the code a branch on sourceforge, and tools to keep their sources current. > Also, I'd like to know if I'm doing anything particularly boneheaded here, > since I'm far from a CVS guru. If you have any CVS questions, just ask. By now I consider myself a CVS guru on how to use CVS (havent studied the internals, or details of the pserver protocol except for what is being used in cvsweb) -- Henrik |