[cvsnt] Re: Missing author bug in 2.0.58 when cvs invoked by postcommit script

[John Kinson]

> The user should never be SYSTEM anyway... it'll be the user who is 
> issuing the command (or whoever the user was overridden as).  I'm not 
> sure how you managed it in 2.0.51... it'll automatically drop into 
> cygwin sshd compatibility if SYSTEM tries to invoke a command and use 
> $USER to work out the real username - that's been true for a long time.
> using this script:
> @echo off
> d:
> cd d:\t
> if exist pc_log goto :end
> echo >pc_log
> cd d:\t\test
> echo foo >>Changelog
> cvs commit -m "xx" Changelog
> cd d:\t
> del pc_log
> :end

Thanks for trying this out.  Did you test it by performing a commit to a 
different file in the same module, with SSPI used as the protocol?

I think I'll need to stick with 2.0.51d until I get to the bottom of this.

I was also caught out by the ACL changes in 2.0.52 - I'd assumed that 
the new command line interface would be backwards-compatible, but it's 
not (in fact, running "cvs chacl John.Kinson:crw filename" caused CVSNT 
to crash for me, which surprised me to say the least!).  I've got a 
number of process model scripts that are now broken by 2.0.52+, but I 
guess that's my bad for assuming "supports many more options" meant 
"supports many more _additional_ options" in the release notes :)

Cheers

JK
-- 
http://www.yellowradio.com/

If technology doesn't seem like magic,
it's probably obsolete.