Re: [Monotone-devel] Cvssync problems

[Christof Petig]

Bill Marquette schrieb:
> On 9/29/05, Brian Downing <address@hidden> wrote:
> 
>>One other comment I have is that it's kind of suboptimal for me to have
>>the "Monotone revision ..... author ..... time ....." as the very first
>>line of the CVS comments, as it uses up the space in ViewCVS output that
>>would otherwise be useful.  For example, see:
>>
>>http://common-lisp.net/cgi-bin/viewcvs.cgi/slime/?cvsroot=slime
>>
>>If I were to start hacking on this with monotone, the "Last log entry"
>>column would be filled with nothing but Monotone revision numbers, which
>>is to me far less useful information than the first line of the
>>developer's log entry.
>>
>>Would it be possible to put the Monotone line at the bottom of the
>>message, instead of the top?

This sounds like a LUA hook would be appropriate. Mentioning original
author, original checkin time and revision (for reference) sounded like
a good default to me.

> I ended up hacking as many of those messages out of my local copy.  I
> use CVS as the mainline repository, with monotone as the development
> repository and I get some rather nasty CVS commit messages because of
> it.  The most annoying of which are the branch propagation ones
> (http://cvstrac.pfsense.com/chngview?cn=6478 for instance) which end
> up telling me nothing about the change (part of that message may be
> due to some of the changelog recursion that I removed).

Some time ago merges only commited with "merge of X and Y" which was
totally meaningless to CVS users. So I started to collect the part of
the history which is unknown to the CVS server. Propagate was not on my
mind when I designed this :-(

Is it correct that propagates should not mention their history. Perhaps
a settable log gathering depth is a good idea.

> What I'd love to see is the ability to add in a message when I do
> monotone cvs_push that describes the huge change that I'm making
> instead of pulling in all the changelog history (or in my case as I've
> removed the recursion logic, just the most recent commit message). 
> Sadly, my C++ skills really really suck...ie, I was at the limit of my
> knowledge just adding a boolean to stop the recursion, else I'd try
> and hack something up myself.

To design a user interface for this feature is not easy (especially
since you might push multiple steps with one command).

    Christof

signature.asc
Description: OpenPGP digital signature