On Mon, 2003-12-29 at 16:16, Garrett Rooney wrote:
> > Hm, that's a legitimate question. I wonder why we are/aren't doing
> > this. The signal handler ought to be able to guarantee that we close
> > the database environment cleanly. Can another developer comment on
> > this?
>
> For the 'svn' process we do catch the signal, it's how the cancellation
> stuff works. Repeated signals may not be caught (I don't think we reset
> the signal handler after it fires, and IIRC some systems require that),
> and in those cases (hiting control-c multiple times) you can still kill
> the process, but if you hit control-c once it should let you exit
> cleanly. If it doesn't that is a bug.
Hm, yah, I can't reproduce the problem on linux:
$ svn co file:///home/sussman/scratch/repos moo
A moo/trunk
A moo/trunk/file.zip
A moo/vendor
A moo/vendor/2.0
A moo/vendor/2.0/file.zip
A moo/vendor/current
A moo/vendor/current/file.zip
A moo/vendor/1.0
subversion/clients/cmdline/main.c:625: (apr_err=200015)
svn: caught SIGINT
$ svn ls file:///home/sussman/scratch/repos
trunk/
vendor/
$
So it seems we are doing the right thing.
>
> For 'svnadmin' i don't think we have a cancellation system in place, so
> if you hit control-c it will most likely mess up the db.
Perhaps we should file an issue on this?
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Mon Dec 29 23:29:41 2003