[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: Ivan Zhakov <chemodax_at_gmail.com>
Date: 2005-11-14 14:51:10 CET

On 11/14/05, Daniel Berlin <dberlin@dberlin.org> wrote:
> On Mon, 2005-11-14 at 08:38 -0500, Daniel Berlin wrote:
> > On Mon, 2005-11-14 at 10:20 +0300, Ivan Zhakov wrote:
> > > On 11/14/05, Daniel Berlin <dberlin@dberlin.org> wrote:
> > > > On Sun, 2005-11-13 at 22:07 +0000, Julian Foad wrote:
> > > > > Greg Hudson wrote:
> > > > > > On Fri, 2005-11-11 at 12:19 -0500, Daniel Berlin wrote:
> > > > > >
> > > > > >>Is it worth the expense of stat'ing every file on every update on some
> > > > > >>common file systems, just to issue an error message *early* about an
> > > > > >>incredibly uncommon case that issues an error message anyway?
> > > > > >
> > > > > > I agree with you; stat'ing every entry is too much of a cost for this.
> > > > >
> > > > > Yes, and your broad plan for fixing it sounds fine. (I haven't looked any closer.)
> > > > >
> > > > > > On a somewhat orthogonal note, the error mesage "Working copy 'foo' is
> > > > > > missing or not locked" is, in my view, a bug any time the user sees it.
> > > > >
> > > > > That's why the default message for that error is "Working copy not locked; this
> > > > > is probably a bug, please report". Unfortunately this particular usage
> > > > > replaces the default message with one that adds some information but loses the
> > > > > second part. That's a bug: it should either include all relevant parts of the
> > > > > default message, or issue its own message in a linked error rather than as a
> > > > > replacement.
> > > >
> > > >
> > > > Well, i've changed it to the attached, which tries to discover more
> > > > interesting reasons for why we couldn't retrieve the adm_access.
> > > >
> > > > I hate having error handling here, but it looks like it is the only
> > > > place lower in the hierarchy to do it, since this is the failure point,
> > > > and the only thing above it is the retrieve call.
> > > >
> > > > Take a look at report_revisions and you'll see what i mean.
> > > Daniel, why you have decided don't change generic error from "Working
> > > copy is missing or not locked" to "Working copy not locked; this is
> > > probably a bug, please report".
> >
> > Uh, i didn't change this at all.
> > I'm not sure what you mean.
>
> Are you saying i should have changed the generic error to say this is a
> bug? If so, i didn't change it because i had no idea that was the old
> generic error :). I'm happy to change it to whatever.
I am saying about remove "is missing" part of message, because we know
that working copy exists.

> >
> > > I consider at this point we know that
> > > working copy exists, right?
> >
> > No, actually, we don't.
> >
>
> Let me follow up here.
>
> The function that fails in any case if svn_wc_adm_retrieve. All we know
> at that point is that we couldn't retrieve an adm_access baton for a
> subpath :)
Yes, but you have additional checks for path existance. So you know
that working copy path exists.

--
Ivan Zhakov
Received on Mon Nov 14 14:54:18 2005

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