Re: svn commit: r1126553 - in /subversion/trunk/subversion: libsvn_wc/upgrade.c tests/cmdline/upgrade_tests.py tests/cmdline/upgrade_tests_data/format_28.tar.bz2

On May 23, 2011 12:48 PM, "Julian Foad" <julian.foad_at_wandisco.com> wrote:
>
> Greg Stein wrote:
> > On May 23, 2011 11:57 AM, <julianfoad_at_apache.org> wrote:
> > >...
> > was 40 hex digits. */
> > > - if (finfo->filetype == APR_REG && strlen(path) == 40)
> > > + if (finfo->filetype == APR_REG
> > > + && (strlen(svn_dirent_basename(abspath, pool))
> > > + ==
> >
> > On my phone, so I can't tell, but ... can basename() take NULL for the
pool?
>
> Yes it can. You thinking we should pass NULL and so squeeze a few more
> microseconds out of the developer-only upgrade scenario?

The basename is completely transient, so there's no reason to alloc and
copy.

That said: you're right. It was a reflexive suggestion, which is kinda moot
given the larger context.

Cheers,
-g
Received on 2011-05-23 19:17:26 CEST