Maintaining a partial cvs workarea

[Axel Thimm]

Hi,

I'm keeping only a partial checkout of the packages, e.g. the ones I'm
maintaining. Now I'd like to be able to cvs up and have all updates
flow in, but if I do so cvs will want to get all other thousand
packages in.

Until now I'm using a poor man's solution with a for loop and
pushd/popd, but it's extremely slow due to login in for each package.

Is there a more clever way to get cvs up running w/o pulling in all of
the cvsroot? I could probably manually edit CVS/Entries, but this feels
a bit dirty. What are other packagers doing?

[Axel Thimm]

Hi,

On Sun, Aug 24, 2008 at 07:44:39PM +0300, Axel Thimm wrote:
> I'm keeping only a partial checkout of the packages, e.g. the ones I'm
> maintaining. Now I'd like to be able to cvs up and have all updates
> flow in, but if I do so cvs will want to get all other thousand
> packages in.
>
> Until now I'm using a poor man's solution with a for loop and
> pushd/popd, but it's extremely slow due to login in for each package.
>
> Is there a more clever way to get cvs up running w/o pulling in all of
> the cvsroot? I could probably manually edit CVS/Entries, but this feels
> a bit dirty. What are other packagers doing?

Sometimes it helps posting a problem to think more about it and solve
it. For posterity and google searches:

Actually what I wanted is already the default. But one usually wants
cvs to automatically discover new folders and pull them in (the -d
option). This is so common, that at one time in your cvs life you will
add it to ~/.cvsrc as a default option like

$ cat ~/.cvsrc 
diff -ud
update -P -d

and you will forget about it, and you will make silly help requests
like the one I did above. :)

Now you dont need to remove the otherwise useful -d switch to update,
instead use

cvs -fq up

to get the desired (non-verbose) updating w/o getting new packages in.

[Jeroen van Meeuwen]

Axel Thimm wrote:
> Hi,
>
> I'm keeping only a partial checkout of the packages, e.g. the ones I'm
> maintaining. Now I'd like to be able to cvs up and have all updates
> flow in, but if I do so cvs will want to get all other thousand
> packages in.
>
> Until now I'm using a poor man's solution with a for loop and
> pushd/popd, but it's extremely slow due to login in for each package.
>
> Is there a more clever way to get cvs up running w/o pulling in all of
> the cvsroot? I could probably manually edit CVS/Entries, but this feels
> a bit dirty. What are other packagers doing?

I do:

mkdir ~/CVS/
cvs co package1
cvs co package2

(just for reference of what is on my computer)

Then, to update all of my checkouts, I do:

cd ~/CVS
cvs up -d

Kind regards,

Jeroen van Meeuwen
-kanarip

[Chris Weyl]

2008/8/24 Axel Thimm :
> Hi,
>
> I'm keeping only a partial checkout of the packages, e.g. the ones I'm
> maintaining. Now I'd like to be able to cvs up and have all updates
> flow in, but if I do so cvs will want to get all other thousand
> packages in.
>
> Until now I'm using a poor man's solution with a for loop and
> pushd/popd, but it's extremely slow due to login in for each package.
>
> Is there a more clever way to get cvs up running w/o pulling in all of
> the cvsroot? I could probably manually edit CVS/Entries, but this feels
> a bit dirty. What are other packagers doing?

With respect to the ssh logins required for each cvs operation, I tend
to use opportunistic connection multiplexing.  e.g., in my
~/.ssh/config I have:

----
ControlMaster auto
ControlPath   ~/.ssh/sockets/%h_%p_%r_multi.sock

Host cvs.fedora.redhat.com
    Compression      yes
    CompressionLevel 3
----

And then I just do a "ssh -f -N cvs.fedora.redhat.com".  It
authenticates me once, then just kicks around in the background until
I perform a network operation though CVS, at which point the "new"
connection is routed through the existing one.  If I haven't forked
off a connection to c.f.r.c, no biggie, ssh just connects per usual.

This won't help with selectively pulling down CVS, but it should make
each operation a smidge faster :-)

-Chris

[Dennis Gilmore]

On Sunday 24 August 2008 04:16:33 pm Chris Weyl wrote:
> 2008/8/24 Axel Thimm :
> > Hi,
> >
> > I'm keeping only a partial checkout of the packages, e.g. the ones I'm
> > maintaining. Now I'd like to be able to cvs up and have all updates
> > flow in, but if I do so cvs will want to get all other thousand
> > packages in.
> >
> > Until now I'm using a poor man's solution with a for loop and
> > pushd/popd, but it's extremely slow due to login in for each package.
> >
> > Is there a more clever way to get cvs up running w/o pulling in all of
> > the cvsroot? I could probably manually edit CVS/Entries, but this feels
> > a bit dirty. What are other packagers doing?
>
> With respect to the ssh logins required for each cvs operation, I tend
> to use opportunistic connection multiplexing.  e.g., in my
> ~/.ssh/config I have:
>
> ControlMaster auto
> ControlPath   ~/.ssh/sockets/%h_%p_%r_multi.sock
>
> Host cvs.fedora.redhat.com
>     Compression      yes
>     CompressionLevel 3
>
> And then I just do a "ssh -f -N cvs.fedora.redhat.com".  It
> authenticates me once, then just kicks around in the background until
> I perform a network operation though CVS, at which point the "new"
> connection is routed through the existing one.  If I haven't forked
> off a connection to c.f.r.c, no biggie, ssh just connects per usual.
>
> This won't help with selectively pulling down CVS, but it should make
> each operation a smidge faster :-)
>                       -Chris

That shouldnt work with the Makefiles.  since they all use 
cvs.fedoraproject.org not the old legacy address :)

Dennis

[Chris Weyl]

2008/8/24 Dennis Gilmore :
> That shouldnt work with the Makefiles.  since they all use
> cvs.fedoraproject.org not the old legacy address :)

s/cvs.fedora.redhat.com/cvs.fedoraproject.org/ in the above then :)
I've never had any issues....  Though, I _really_ ought to update my
checkout to use the new addy.

(Yes, I've had my cvs checkout since way before it became an old
legacy address, and I hate change :-) )

-Chris

[Axel Thimm]

On Sun, Aug 24, 2008 at 02:55:41PM -0700, Chris Weyl wrote:
> 2008/8/24 Dennis Gilmore :
> > That shouldnt work with the Makefiles.  since they all use
> > cvs.fedoraproject.org not the old legacy address :)
>
> s/cvs.fedora.redhat.com/cvs.fedoraproject.org/ in the above then :)
> I've never had any issues....  Though, I _really_ ought to update my
> checkout to use the new addy.
>
> (Yes, I've had my cvs checkout since way before it became an old
> legacy address, and I hate change :-) )

I also need to move my Roots to something new age. Wasn't there a tool
for CVS? Or some perl/sed magic to do that?