svn commit: r356695 - head/usr.bin/diff

[Mark Johnston]

On Tue, Jan 14, 2020 at 08:05:21AM -0600, Kyle Evans wrote:
> On Tue, Jan 14, 2020 at 6:12 AM Li-Wen Hsu <lwhsu at freebsd.org> wrote:
> >
> > On Tue, Jan 14, 2020 at 2:29 AM Mark Johnston <markj at freebsd.org> wrote:
> > >
> > > Author: markj
> > > Date: Mon Jan 13 18:29:47 2020
> > > New Revision: 356695
> > > URL: https://svnweb.freebsd.org/changeset/base/356695
> > >
> > > Log:
> > >   Optimize diff -q.
> > >
> > >   Once we know whether the files differ, we don't need to do any further
> > >   work.
> > >
> > >   PR:           242828
> > >   Submitted by: fehmi noyan isi <fnoyanisi at yahoo.com> (original version)
> > >   Reviewed by:  bapt, kevans
> > >   MFC after:    1 week
> > >   Differential Revision:        https://reviews.freebsd.org/D23152
> > >
> > > Modified:
> > >   head/usr.bin/diff/diffreg.c
> > >
> > > Modified: head/usr.bin/diff/diffreg.c
> > > ==============================================================================
> > > --- head/usr.bin/diff/diffreg.c Mon Jan 13 18:26:27 2020        (r356694)
> > > +++ head/usr.bin/diff/diffreg.c Mon Jan 13 18:29:47 2020        (r356695)
> > > @@ -349,6 +349,11 @@ diffreg(char *file1, char *file2, int flags, int capsi
> > >                 goto closem;
> > >         }
> > >
> > > +       if (diff_format == D_BRIEF) {
> > > +               rval = D_DIFFER;
> > > +               status |= 1;
> > > +               goto closem;
> > > +       }
> > >         if ((flags & D_FORCEASCII) == 0 &&
> > >             (!asciifile(f1) || !asciifile(f2))) {
> > >                 rval = D_BINARY;
> >
> > Hi Mark,
> >
> > This one seems breaks test case usr.sbin.etcupdate.fbsdid_test.main:
> >
> > https://ci.freebsd.org/job/FreeBSD-head-amd64-test/13920/testReport/junit/usr.sbin.etcupdate/fbsdid_test/main/
> >
> > Revert this can get the case pass. Can you check if the code or the
> > test needs to be fixed?
> >
> 
> Whoops... that was a bit of an oversight. We can only short-circuit if
> ignore_pats == NULL, otherwise we need to regress back to doing
> line-by-line comparison. That should still see a benefit for the
> majority of cases, though, I would think.

Thanks, I tested and committed that change.