Bug #43289 for App-SVN-Bisect: If all revisions between start and end are skipped, svn-bisect arbitrarily picks a revision

Sat Feb 14 15:18:02 2009 greenrd [...] greenrd.org - Ticket created

Subject:

If all revisions between start and end are skipped, svn-bisect arbitrarily picks a revision

Using svn-bisect, I had narrowed down a bug to a range of revisions. Revision N (the exact value doesn't matter) didn't have the bug, revision N+7 did, but revisions N+1..N+6 were skipped because they couldn't be tested. After I performed the last skip operation, svn-bisect arbitrarily decided that the bug was introduced in revision N+7. But actually, the bug *could* have been introduced in any one of N+1..N+7, and I think the final output of svn-bisect should reflect that fact.

Sat Feb 14 19:30:28 2009 INFINOID [...] cpan.org - Correspondence added

Hi! Thanks for the report. Okay, I think I understand what you're trying to say. When you skip a revision, svn-bisect assumes you're telling it totally ignore it, but the resulting message may not be totally accurate. So, what do you think it should say? "The first known bad revision is rNNN, the change may have occurred there or in one of the N skipped revisions immediately preceding it"? It's a bit verbose, but I wouldn't object to having it say something along those lines.

Sat Feb 14 19:30:28 2009 The RT System itself - Status changed from 'new' to 'open'

Sat Feb 14 19:31:47 2009 INFINOID [...] cpan.org - Taken

Sun Feb 15 06:10:44 2009 greenrd [...] greenrd.org - Correspondence added

On Sat Feb 14 19:30:28 2009, INFINOID wrote: Show quoted text

> Okay, I think I understand what you're trying to say. When you skip a > revision, svn-bisect assumes you're telling it totally ignore it, but > the resulting message may not be totally accurate. So, what do you > think it should say? > > "The first known bad revision is rNNN, the change may have occurred > there or in one of the N skipped revisions immediately preceding it"?

Yes, that's what I mean.

Sat May 02 09:24:14 2009 INFINOID [...] cpan.org - Correspondence added

On Sun Feb 15 06:10:44 2009, greenrd wrote: Show quoted text

> On Sat Feb 14 19:30:28 2009, INFINOID wrote:

> > Okay, I think I understand what you're trying to say. When you skip a > > revision, svn-bisect assumes you're telling it totally ignore it, but > > the resulting message may not be totally accurate. So, what do you > > think it should say? > > > > "The first known bad revision is rNNN, the change may have occurred > > there or in one of the N skipped revisions immediately preceding it"?

> > Yes, that's what I mean.

svn-bisect version 0.8 is on its way to CPAN with a fix for this. Thanks!

Sat May 02 09:24:15 2009 INFINOID [...] cpan.org - Status changed from 'open' to 'resolved'