[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: The cost of svn_io_get_dirents2 and early error messages

From: Daniel Berlin <dberlin_at_dberlin.org>
Date: 2005-11-14 02:36:17 CET

On Sun, 2005-11-13 at 17:19 -0800, Greg Stein wrote:
> On Fri, Nov 11, 2005 at 12:19:17PM -0500, Daniel Berlin wrote:
> >...
> > This would mean they would be issued a bit later. In the
> > file->directory case, it wouldn't happen until you received updates to
> > that file.
>
> What if you never received updates?

*Right now* (IE without any patches), nothing will give you any
indication that this is an error.

dberlin@linux:/mnt/gccstuff/gcc-dataflow>
~/svn-trunk/subversion/clients/cmdline/svn update config.guess
At revision 106878.
dberlin@linux:/mnt/gccstuff/gcc-dataflow> rm config.guess
dberlin@linux:/mnt/gccstuff/gcc-dataflow> mkdir config.guess
dberlin@linux:/mnt/gccstuff/gcc-dataflow>
~/svn-trunk/subversion/clients/cmdline/svn update config.guess
At revision 106878.
dberlin@linux:/mnt/gccstuff/gcc-dataflow>

In fact, diff won't even complain either.

> For example, svn's /trunk/COPYING
> is a couple years old. If that was replaced with a directory, then I'd
> want to have some kind of signal about it "soon", rather than several
> years later when somebody tweaks the license.

svn status would show something weird right now if you do this

dberlin@linux:/mnt/gccstuff/gcc-dataflow>
~/svn-status/subversion/clients/cmdline/svn status config.guess
svn: Working copy 'config.guess' is missing or not locked

Actually, with the patch i attached, it will say:

dberlin@linux:/mnt/gccstuff/gcc-dataflow>
~/svn-propcaching/subversion/clients/cmdline/svn status config.guess
svn: Expected config.guess to be a file but found a directory

which is much more informative.

If you just use status with no command line arguments anything else, it
will show "~" next to config.guess both now and with the patch
attached.

However, doing this during updates would, IMHO, be really wrong. We
don't do it right now, and ideally, the updater wouldn't have to touch
*anything* that isn't actually updated.
So expecting mass verification to happen there seems, uh, not the place
to do it. I'm fine with doing entry vs stat verification during
operations that need to stat everything anyway (like diff, for example).

Considering this is not some "typical thing users do every day", and it
was presumably done by you anyway, i'm fine with it not doing it on
update.

--Dan

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Nov 14 02:37:40 2005

This is an archived mail posted to the Subversion Dev mailing list.