[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: 'Dying gasps received from client.' with CVS 1.12.11
From: |
Joerg Bullmann |
Subject: |
Re: 'Dying gasps received from client.' with CVS 1.12.11 |
Date: |
Wed, 19 Jan 2005 13:22:58 +0100 (CET) |
Hi Mark and Thomas,
> > The zlib library went from 1.1.4 to 1.2.1 on 2004-11-11. There has been
> > one other compression complaint, but I am not sure if Derek ever figured
> > out what was wrong or not.
I have tried to find out what is going wrong here myself
in MacCVSClient which links against Mac OS X zlib version
(which according to the lib name seems to be 1.1.3 !!).
Did not manage to find the problem yet, though. Have tried
to link a 1.2.1 that I have built myself here. To no avail
either.
> OK, this can be the reason for the 'dying gasps'.
>
> >>How do you think, the client should detect, whether the paths are
> >>relative? Please note, that CVSNT might return absolute paths without
> >>a leading slash, e.g.
> >>...
> >>Clear-static-directory ./
> >>c:\repository/test/
> >>...
> >>or
> >>Clear-static-directory ./
> >>c:/repository/test/
> >>...
In MacCVSClient I do it this way: I check whether the CVS root is
a prefix of the path. If that is the case, I know I have an absolute
path, otherwise a relative one.
I think this is OK because the CVS root is an absolute path itself
(either starts with / or a "X:" drive prefix). It can thus never be
a relative path and I can be sure that once I detect it as a prefix
of a path this path is absolute.
Am I making a mistake here? Anyway, so far this has worked here. It
is still in pre-release code only.
> That's bad, because SmartCVS code needs to support CVSNT and GNU CVS and
> currently I have no clue, how to support GNU CVS >= 1.12.10. Maybe you
> rethink about the relative paths or send them only, when the client
> tells that it supports that?
Hope the above thoughts help.
Cheers,
Joerg