Re: [Gnu-arch-users] Problem with cscvs

[Mirian Crzig Lennox]

address@hidden (Nuno Ferreira) writes:
> On Qua, 2004-01-14 at 17:33 +0000, Nuno Ferreira wrote:
>> On Qua, 2004-01-14 at 15:20, Charles Duffy wrote:
>> > Please let me know if the attached patch works for you.
>> 
>> Sorry, it doesn't. Now it fails with:
>> A doc/userguide/images/11-1.png [1.1]
>> cvs update: cannot open CVS/Entries for reading: No such file or directory
>> cvs [update aborted]: no repository
>> Performing undo due to exception: "cvs -q update -p -r1.1 
>> 'doc/userguide/images/11-1.png'" failed with exit status 256
>> 
>> The directory is created but cvs command still fails because no CVS
>> sub-directory exists.

For what it's worth, I was getting that error a LOT when using cscvs.
However, I recently updated to cscvs--experimental--1.1--patch-77 and
the problems have all gone away.

One detail easily missed is that the
I-Just-Want-To-Use-It-To-Sync-With-Arch instructions in the README
have altered slightly.  For example, the initial import is done "by
hand", not with cscvs -i as previously.

I have found that by following these updated instructions precisely, I
have been able to make cscvs process successfully a substantial CVS
repository with nigh on five years of edit history.  Quite impressive!

One small point: it would be nice if cscvs had a simple way of getting
a list of every filename path which appears in a given range of
patchsets.  (Perhaps it does and I missed it).  The method I came up
with was 

     cscvs log -P<branch> | awk '/^\t/ { print $1; }' | sort | uniq

which works if it can be assumed that no line in the log messages will
begin with a tab character.  (Of course, the nice thing about cscvs
using sqlite is that you can fairly efficiently query the patchset
information directly.)

That notwithstanding, cscvs is a wonderful tool, and I'm ever so
grateful to Charles and others for their efforts.

cheers,
Mirian