fake merges (was: [PATCHES] Bootstrap: Allow user overriding of $AUTOCONF and $PERL.)

[Ralf Wildenhues]

* Stefano Lattarini wrote on Sun, Aug 08, 2010 at 01:13:07PM CEST:
> At Sunday 08 August 2010, Ralf Wildenhues wrote:
> > * Stefano Lattarini wrote on Sun, Aug 08, 2010 at 12:53:21PM CEST:
> > > At Sunday 08 August 2010, Ralf Wildenhues wrote:
> > > > Also, regenerating the tree with Autoconf 2.67 and committing
> > > > that separately is preapproved for maint.
> > > 
> > > Couldn't this cause problems when later merging to master?  If
> > > yes, how would those be better solved?
> > 
> > This is solved most easily by having one commit in maint that
> > regenerates the tree, and merging that in the active branches.
> > You may want to do a fake merge in this case to force a correctly
> > created configure script for the branches ...
> Pardon my ignorance, but what do you mean by "a fake merge"?

When you merge a branch A into B, then it can happen that git sees no
conflicts, but in reality things break anyway.  This can easily happen
when, say, A renames a variable, B adds new code using the old variable.
Then you need to fix up the code after the merge.  You can 'git commit
--amend' the fixup to the merge.  I think this is what is called a fake
merge.

In the case of regenerating with newer Autoconf, I just ran bootstrap on
maint with 2.67, merged that into branch-1.11.  That caused a conflict,
but even if it hadn't, I would've rerun bootstrap and commit --amend'ed
so that files are correct for the branch.  Same for master.

I just did that and pushed.

Cheers,
Ralf