[Monotone-devel] Re: What's the plans for cvs_import?

[Richard Levitte - VMS Whacker]

[ I've started using the address I want for free software development,
  address@hidden  I used address@hidden previously.
  Please don't let this confuse you ]

In message <address@hidden> on Fri, 30 Jul 2004 12:14:37 -0400, "graydon hoare" 
<address@hidden> said:

graydon> On Thu, 29 Jul 2004 22:01:42 +0200 (CEST), Richard Levitte - VMS 
Whacker <address@hidden> wrote:
graydon> 
graydon> > It also seems to skip some steps in the history, maybe
graydon> > because the log messages are extremely similar (excatly the
graydon> > same).  However, considering every change happens with a
graydon> > time difference of about 5 minutes, it shouldn't be that
graydon> > hard to tell the difference.
graydon> 
graydon> the default time window is 3 hours (constants.cc, cvs_window).
graydon> this is conservatively high due to the fact that authors and
graydon> changelogs typically change on different checkins, and that
graydon> some trees (eg. gcc accessed over a slow modem with high lock
graydon> contention) have relatively lengthy checkin skew.

I think that what I find most disturbing is that it thinks two commits
*on* *the* *same* *file* is really one commit when the time difference
is less than cvs_window.  I have a very hard time understanding how
two commits on the same file can ever be regarded as one commit.

I'm currently looking more closely at the way
cvs_key::similar_enough() is used, and I have a hard time seeing how
it's used to figure out if *different* files are commited at the same
time or not.  I'll do some experiments and see what I can figure out.

-----
Please consider sponsoring my work on free software.
See http://www.free.lp.se/sponsoring.html for details.

-- 
Richard Levitte                         address@hidden
                                        http://richard.levitte.org/