Re: libgcj-classpath-20060719 merge

[Mark Wielaard <mark at klomp dot org>]

Hi,

On Mon, 2006-07-31 at 14:09 -0600, Tom Tromey wrote:
> >>>>> "Haren" == Haren Visavadia <themis_hv@yahoo.co.uk> writes:
> 
> >> It seems weird that in order to do a merge we must create a third
> >> branch.
> 
> Haren> It makes it into more logical steps:
> Haren> classpath -> gcj-classpath-integration -> trunk
> 
> I don't agree that it is more logical.  I think that doing a merge and
> then fixing it to build correctly is a "normal" operation, one which
> ought not require an intermediate staging branch in order to get a
> good diff.

Indeed. It is infuriating that svn seems to prevent creating the needed
diff unless you first commit. The whole point of creating the diff is to
show it to others so it can be inspected before you commit things to the
server repository.

> That said I don't think it is substantially "less logical" to get
> diffs against the merge-ee base revision.  It is weird for me, and I
> can't really imagine why you'd ever want this, but I'm assuming this
> is just an imagination failure on my part.

I am also not imaginative enough, but I guess we are stuck with it. We
will have to create a new branch to which we actually commit things so
we can create the patches more easily for review and then actually merge
it over to the trunk. If everybody is OK with doing such things.

I am not really happy that I need to commit to the server repository
first and then ask for people to help me spot any errors. I much rather
not change anything on the server till I am convinced the patch/merge is
actually good. But it seems that is how things are now.

I will probably not do that for the final 0.92 merge yet though since I
basically have that done already (except for the last few patches on the
0.92-release-branch, which I seem to keep promising will turn into the
actual release "any day now"...).

Cheers,

Mark