Re: [PATCH 1/6] : Commonise various target-descriptions.c functions

[Yao Qi <qiyaoltc at gmail dot com>]

Alan Hayward <Alan.Hayward@arm.com> writes:

> Did you want me to change the ordering of the patches in the series or did you
> want me to reduce the amount of code that gets moved into arch/ ?
>

Both.

> When I wrote the whole patch series, I only moved across the functions I needed
> from gdb to arch/. It turns out that I needed to move a lot of
> functions. Without
> complicating the code, I think I need everything I moved into arch/
>
> After I had finished writing my code, I looked at the ways of
> splitting it up into a
> series. The simplest method was to put all the moving of functions
> into the first
> patch. I agree that this makes the first patch a little odd to
> review. My reasoning
> was to keep changes to the moved code separate from the moving.
>

For some big patch series, the patches (order and contents) I posted for
review are different from the patches I wrote to get things done.  After
I get my branch working, I'll restart a new branch, to rewrite them
in a way which is more friendly for review/upstreaming, because you've
never know what changes are needed until you finish the work.  It
happens to me that I realize that I need to refactor some part after I
committed ten patches in my branch, so I have to stop, rebase my
patches, to put refactor patches first.

> I can look at making a new version of 2/6 the first patch, and then follow with
> smaller versions of this patch? The final change across the whole series would
> be the same. 

That will be very helpful.  Steps matter, take a look at page 6 of my
presentation,
https://gcc.gnu.org/wiki/cauldron2013?action=AttachFile&do=get&target=port-gdb-tic6x-qi.pdf

-- 
Yao (齐尧)