[cvsnt] Re: Missing author bug in 2.0.58 when cvs invoked by postcommit script
cvs at yellowradio.com
Tue Oct 5 14:43:25 BST 2004
>> Any chance this could be fixed please? I assume it's still acceptable
>> for a postcommit trigger script to use cvs to perform an additional
> It's not a good idea as the script will be called again and you can get
> into a loop.
True, but my script handles that - that's not a problem for me.
> I don't see how postcommit/postcommand can cause that - in fact the
> author field has absolutely no link with anything to do with postcommit.
> It's also not optional... the very first thing written to a new
> revision is the author and will never be 'missing'.
Here's the head of the ,v file after my postcommit script has created
comment @# @;
date 2004.10.05.11.46.31; author ; state Exp;
date 2004.10.05.11.39.21; author SYSTEM; state Exp;
As you can see, revision 1.12 has no author, which is clearly a bug.
The cvsnt server process should never write a ,v file with an
> I suspect this is unrelated to postcommit and something else is going
> on. Try removing the postcommit and committing a few things. Do you
> have any AV on the server?
If I rollback the server to 2.0.51, the script works - the author for
the created revision is SYSTEM. If I upgrade the server to 2.0.58, my
script stops working. Given that 2.0.52 introduced changes to the SSPI
domain processing (usernames are recorded as @DOMAIN\username@, rather
than @username@), isn't this a likely source of the bug when "cvs
commit" is invoked as SYSTEM?
If technology doesn't seem like magic,
it's probably obsolete.
More information about the cvsnt