Quoting "C. Michael Pilato" <cmpilato@collab.net>:
> Branko ÄŒibej <brane@xbc.nu> writes:
>
> > There are no conceptual problems that I see, if that's what you
> > mean. But I really, really believe this is not something we should
> > leave to scripting, and if we can avoid platform-specific code, so
> > much the better.
>
> My thoughts are similar to Branko's. Compiled-in paths are for the
> birds, and pretty much kill most types of binary distribution.
> Searching for a utility in a PATH environment is almost as bad --
> /usr/local/bin (a common Subversion installation location for Unix
> systems) isn't always in the default path for users like httpd.
>
> The right solution is to fork a new process or thread to take care of
> this task.
We can't do it in a thread because of httpd issues, but if
APR_HAS_FORK, we can fork off a process (and really, on Unix
that would probably be the best (simplest?) solution).
Hm, of course, we could _tell_ the repos layer if spawning a thread
is acceptable. Then mod_dav_svn would fork a process, svnserve could
use a worker thread and queue, svn on ra_local could look at argv[0]
and spawn "svnadmin deltify" from there...
Lots of options, really.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Nov 17 16:07:42 2003