DjangoBot: #django-dev

So is there a git user in the house who wants to answer a stupid question from a recent convert?

converting to git? nice

brosner: yeah, hgsvn has started to piss me off, and git svn dcommit actually works as advertised...

jacobkm: yay!

scnr

Anyway, I suspect there's a simple answer here, but: So I've got a series of commits (about a dozen, not that it matters) on a branch. I'd like to apply all those changes to master in a single commit such that when I dcommit it appears as a single commit in SVN.

Ideally I'd like a "git svn commit --all-at-once" but that doesn't seem to exist.

in master... git merge --no-commit --squash the_other_branch

Grr, I know it'd be simple :P

that will not commit locally. then commit it all at once :)

I still don't completely like git, but mostly because I'm afraid it's smarter than I am, and being shown to be stupider by a bunch of perl makes me unhappy.

yikes

heh, you'll get over it once it saves your time

:)

So far I want to have babies with git commit --amend.

Other than that I'm mostly using it as a way of staging SVN commits, which is fine enough.

its the ones you can rename at will ;)

Honestly, SVN doesn't really have to go very far to handle 90% of what most mortals use DVCS for, I think.

#7125 is getting no love ;)

Ping on #6168

Can someone please just check this in for me? :P I'm tired of patching my copy of Django lol

I think nobody's running into the bug since nobody's changing model stuff very much...

ericflo: i'd commit it if i had trunk access ;)

brosner: Thanks, the thought counts :)

brosner: I'm almost done with my patch-updating. Not much changing except a lot of var == None turning into var is None

(Malcolm told me the latter is better, and I'm inclined to implicitly trust him)

yeah there were a couple of very minor things i saw

the kwarg on the create many to many manager function (i forget the exact name) probably should use None instead of False for the default value to be consistent

ericflo: Yeah you are supposed to do that because all instances of None have the same id() (which is what the `is` tests I thnk)

What ticket are we talking about?

and not to be very pick imo instead of del Something.add maybe delattr(Something, "add") seems cleaner in the latter. but really doesn't matter

Alex_Gaynor: #6095

also is there a reason why create is being deleted differently than add and remove?

Oh, intermediary :P

brosner: Yeah, it breaks when you do del.

brosner: Since there's already an entry in the method lookup table, inherited.

ah

brosner: If you delete the attribute on the object, it just goes upstream.

then you might want to revise the comment by the dels

brosner: I did locally :)

brosner: I caught that and was like, wtf, why did I do that, took a few tests to figure it out lol

ah gotcha. a comment there might be nice

brosner: good point, will do

jkocherhans: in regards to nfa save_as support how do you think unique=True should be handled? it would make model-aware validation a blocker to mirror trunk's behavior.

brosner: would it? from what I remember, trunk will give you a db error right now if a unique constraint fails... I haven't used the admin in quite awhile now though, I could be wrong.

trunk gives the proper validation error

i remember testing it and it just said the page couldn't be added because in my case the slug was already in use

i should really test it a bit more ;)

k. I guess the only other option would be to write that validation into the save view, which is sub-optimal, but no idea when model validation will be done

I don't think Honza has had a lot of time to work on it since PyCon

maybe it's unique_together that raises a db error

unique raises a DB error I think

yeah. id be in favor to let the exception happen and let the user ensure the form has the right validation. id hate to have something like that block a merge to be honest.

IntegrityError

Alex_Gaynor: not on trunk's admin

Hrm, really? I thought it was

If the underlying DB has an index of the unique_together fields, a DB API exception ought to surface, even in trunk..

but in oldforms manipulators it has a validator for a fk with unique

Well whats happening with ModelValidation

unique_together might be failing, i never tested that ;)

Alex_Gaynor: Honza

err

Honza's work?