[PATCH 0 of 2] log: slowpath

[Matt Mackall]

[working through backlog]

On Mon, 2010-07-05 at 22:32 +0900, Nicolas Dumazet wrote:
> 2010/7/5 Greg Ward <greg-hg at gerg.ca>:
> > On Sun, Jul 4, 2010 at 10:20 AM, Nicolas Dumazet <nicdumz at gmail.com> wrote:
> >> 2) --removed option apparently had never been implemented. Ever.
> >>   I understand it as "only include revisions where fieles were removed". Does
> >>   it makes sense in this way, and can you think of any other meaning?
> >>   If we agree on the meaning, log docstring probably has to be improved to
> >>   include "only"/or be clearer.
> >
> > I'm not familiar with the code, but --removed *does* have an effect.
> > Specifically, it shows the entire history of the file *including* the
> > changeset that removed it.  My understanding is that it does this by
> > reading the manifest log, not just the file log, which is why it's so
> > much slower.
> 
> Yesss of course. Now that you say this, it makes perfect sense.
> Patch #2 should obviously be discarded then.

That's incorrect - it looks at the files field of each changeset for
matches. That will include deletes as well as 'twins' - sibling
revisions that have the same hash and are stored as one revision in the
filelog.

-- 
Mathematics is the supreme nostalgia of our time.