On Oct 19, 2006, at 20:01, Talden wrote:
>> On Oct 19, 2006, at 18:47, Talden wrote:
>> > Consider that with 1000 devs, committing once a week each you've
>> got
>> > something like one commit every 3 minutes in an 8-hour day. You
>> can't
>> > commit without updating so you update - an update would
>> therefore need
>> > to be guarenteed to complete inside 3 minutes.
>
> On 10/20/06, Ryan Schmidt wrote:
>> Not true. You most certainly can commit if you are not at the HEAD
>> revision. You only need to be at the HEAD revision of the files and
>> directories you're currently committing. Of course, if 1000
>> developers are all committing the same files all the time, then you
>> might run into more conflicts than you would like, but I'm not sure
>> that's a realistic scenario.
>
> Being at the head revision requires updating right. Sure you're not
> updating the entire working copy (just the subtree that encompasses
> the full scope of the commit).
>
> So you finish your task, you update, you test, you go to commit.
> Whoops gotta update. Since you only have 3 minutes on average before
> the HEAD advances again, there's no way you can update and do any
> reasonable checking, you just hope there are no adverse interactions
> and commit.
Again, you only have this problem if someone modifies the exact same
files you're commiting within those 3 minutes.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Fri Oct 20 04:55:46 2006