status messages during hg clone

[Adrian Buehlmann]

On 22.03.2008 10:31, Dirkjan Ochtman wrote:
> On Sat, Mar 22, 2008 at 1:37 AM, Adrian Buehlmann <adrian at cadifra.com> wrote:
>>  Basically, what I don't exactly understand is, why you didn't
>>  just merge bf7afddcdca2 (my change) and d2713d902524 (crew main branch).
> 
> Right. In hindsight, that might've been better. I based my behavior on
> how I interacted with crew before when we were discussing one of my
> patches in IRC. They would usually just rebase my patch so that it
> wouldn't need a merge (or ask me to rebase it). That's a bit
> different, though, since there was more interaction.

The point that made me thinking here was: what would be so bad about a
Mercurial merge here? Why are people trying to avoid Mercurial merges?

There wouldn't have been any actual *files* to merge. I fear the term
"to merge a head" does have some (potentially) wrong connotations.

>>  Now, I could use that discussed new feature "closing branches"
>>  to mark bf7afddcdca2 as closed or strip it away. But stripping
>>  it would break the link in my archived posting (ok, no one would
>>  probably be throughly disappointed here...).
> 
> I always keep my patches in mq on top of a repo, so I can easily
> rebase a patch or pop it in favor of an upstream fix.

Ok. But why was rebasing needed here? Why is rebasing needed in
general? A Mercurial merge changeset would actually record the
answer to that "who applied that patch to crew repo?" question.

>>  As a side note: I choose not to send my change as a patch to
>>  the list, because I thought sending 35 KB text to the list
>>  would have been bad.
> 
> I think sending a 35 KB patch would've been just fine. Makes review a
> little easier.

Ok. Next time I will send up to 35 KB to the list. If somebody complains,
I can say you told me :-)

>>  My thinking was that I could push potential future changes there
>>  as well, so that people could review them using the browser.
> 
> It is still useful to have a public mirror with patches applied where
> people can review and/or pull from.

You mean my public repo clone
(http://www.cadifra.com/cgi-bin/repos/mercurial-abuehl/) is still
useful? That's great! The thing to keep in mind for me is to
always base my changes I push there on a clean changeset from crew.
But I actually might push a series of changesets for a bigger "patch"
to that repo if it's a larger change (very unlikely to happen anyway).

If such a series would be accepted, the smaller steps towards
the bigger change would still be visible once pulled into crew.
That's something I like a lot of Mercurial, as larger changes sometimes
contain a few refactorings which should best not be intermixed into the
same changeset with externally visible program behaviour changes.

Still somewhat unanswered here is what I should do with my stale head
bf7afddcdca2 over there now. For example, merging that into the main line
-- as hinted by the section "pruning dead branches" from TipsAndTricks on
the wiki -- would be a bad idea as that would just create yet another
"unclean" branch (a branch diverged from crew).

So, only options left are stripping ("hg strip" from MQ extension, see
"hg help mq") and simply leaving that head as it is (or marking it as
"closed" as soon as the "closing branches" feature is available).