I could make git-cl track what your patch was last uploaded against,
and warn on dcommit if that upload was not based on an upstream svn
commit.  That would be pretty easy:
 - on upload, stuff the diff base into a property of the branch
 - on dcommit, call something like "git cat-file -p DIFFBASE | egrep
-q ^git-svn-id:"

I seem to recall Paweł sending me a patch to do something like #1 but
it's probably buried in my inbox somewhere... :\

On Thu, May 21, 2009 at 10:34 AM, Marc-Antoine Ruel <[email protected]> wrote:
> [+chromium-dev]
> Well, the plain old:
> If "diff of stuff being committed" != "diff of codereview"
>   Error -> "You are trying to commit code that is different from what you
> reviewed."
> Note that this applies to svn too. I think I could implement that on a
> PRESUBMIT script. I'm just afraid it'd slow down the commit. :/  One thing
> to help would be to cache the diff on upload.
> M-A
> On Wed, May 20, 2009 at 12:44 PM, Evan Martin <[email protected]> wrote:
>>
>> I would like this as a feature, since I do it too.
>>
>> The technical implementation is a bit tricky, since git doesn't really
>> know how branches are related.  They're just pointers into the commit
>> graph.  You could do something like "on commit, verify that no other
>> branch is a subset of this one" but that might be fragile and won't
>> catch the case where you have something like
>>
>>  A---B   <- review1
>>        \
>>         C---D---E   <- review2
>>
>> and then you do a new commit on review1 due to review feedback, and
>> then try to dcommit review2, and review1 is no longer a strict subset
>> of review2.
>>
>> On Wed, May 20, 2009 at 6:15 AM, Marc-Antoine Ruel <[email protected]>
>> wrote:
>> > Just askin', would it be possible when a code review (branch) is based
>> > on another code review (another branch) to block the check-in of the
>> > dependent code review until the root code review is checked-in?
>> >
>> > Because I messed up and checked-in the dependent change without
>> > checking-in the initial change, the single commit included both
>> > changes. Not a big deal but still not cool even if it's 100% my fault.
>> >
>> > I guess I could implement that as a git hook but would it be useful to
>> > have it in git-cl? Maybe not, what do you think?
>> >
>> > M-A
>> >
>
>

--~--~---------~--~----~------------~-------~--~----~
Chromium Developers mailing list: [email protected] 
View archives, change email options, or unsubscribe: 
    http://groups.google.com/group/chromium-dev
-~----------~----~----~----~------~----~------~--~---

Reply via email to