[webkit-dev] spinoff from webkit-patch -g discussion

[Dirk Pranke]

On Wed, Feb 29, 2012 at 3:06 PM, Shawn Singh <shawnsingh at chromium.org> wrote:
> Quote from the discussion:
>
>>
>> that's why I said that having something like webkit-patch upload
>> range_of_commits will be nice to have, as you wouldn't have to create a
>> new branch and rebase several commits, just to upload a new patch to the
>> bz.
>
>
>> You can do this with the current -g option by adding a commit range, e.g.
>> -g=commit1..commit2. AFAIK, the only thing you can't do currently with -g is
>> pass a commit range *and* include the staged/working copy changes.
>> Under the hood it basically does what you described (create a new branch,
>> copy the commits over as a single commit, upload from the branch, etc).
>
> I didn't want to hikack the other thread, but I wanted to discuss this
> -git-commit flag.
>
> It makes sense that we can't commit a commit range + unstaged changes, and
> its easy enough to commit before uploading.  However, one problem I've
> encountered is that webkit-patch may make changes to the changelogs, in
> particular the "reviewed by" clause. But then, because its an unstaged
> change, it uploads the "reviewed by nobody" clause.  So I have to manually
> edit the "reviewed by..." statement, commit, and then upload.
>
> Am I missing something simple?  It seems to defeat the purpose of
> automatically editing the changelogs, if I still have to quit and commit it
> myself before uploading again.
>

No, you are not missing something; this is actually one of the main
reasons I started the thread.

> Would it be reasonable to force an error to upload a commit range, when
> there are unstaged changes?  I've observed this only for land-safely and
> upload.  I've been unwilling to experiment with landing directly because I
> don't want it to accidentally commit a "reviewed by nobody" to the
> changelogs.
>

In my view, I would actually rather upload the combination of
committed + staged + unstaged changes rather than be told I have to
commit things; in other words, I actually prefer to commit what I've
uploaded rather than upload what I've committed.

landing directly shouldn't be an an issue, insofar as webkit-patch
will stop (or at least warn) you if you still have an OOPS in the
Changelog.

-- Dirk