This is the mail archive of the
mailing list for the GCC project.
Re: gcc-ss-20001113 is now available
- To: gcc at gcc dot gnu dot org
- Subject: Re: gcc-ss-20001113 is now available
- From: Jason Molenda <jason-gcc-list at molenda dot com>
- Date: Wed, 15 Nov 2000 16:26:53 -0800
- Cc: tej at melbpc dot org dot au
Tim Josling wrote:
> My cvs update for the weekly snapshot suddenly failed this week.
As others have already noted, anoncvs.cygnus.com and gcc.gnu.org are
aliases for one another. And the CVS pserver protocol doesn't give CVS
enough information for the hostname to make any difference.
More importantly, the CVSROOT environment variable is ignored when
doing a cvs update operation in an existing tree. The CVS/Root
files will override the CVSROOT environment variable. (and the -d
command line option to cvs will override both of these)
If I had to make a guess, I'd wager that your cvs update failed because
gcc.gnu.org limits pserver access when the load average goes above 15.
I have no idea if the load average has been above 15 recently, but it's
either that or a network outage. If it was the load limiter, you would
have gotten an error message from cvs explaining why the cvs command
Tim also wrote:
> Using the new URL it spent a long time doing nothing at all
> (renaming file X as it is in the way, then do no updates). So I
> am now doing a full refresh :-(.
A cvs update can be slow for a variety of reasons. It make _look_
like it is hung when it is actually waiting for the server to check
the revisions of all of the files in the repository and determine
what needs to be updated. A lot of this client-side checking and
server-side checking takes place before any files are sent.
If cvs was in the process of sending the file differences (it's
pretty obvious - you'll see the "P", "U", etc., update lines), then
a slowdown may be due to network bandwidth. If that's the case,
make sure you're compressing your cvs communications either by
having "cvs -q -z9" in your $HOME/.cvsrc or by running "cvs -q -z9
update" (the -q is unrelated, but often preferred).
Checking out an entirely new tree skips all of the version checks
(because you have no existing files, so all must be sent to you)
so things will start happening immediately, but it will take a long
time to send all of the files over the Internet. Again, compressing
the communications is essential, but doing an update is preferred
even more -- you'll only be pulling over a fraction of the data.
If you're ever in doubt about your cvs operation, try running it
with a '-t' command line option. With a recent cvs version, it
should show you all of the cvs commands that are being sent between
the server and the client. It's an awful lot of output, but you can
see what is really going on.