[illumos-Developer] Fwd: [Urgend] Round eleven of code review for ksh93-integration+POSIX utility update3 ...

Garrett D'Amore garrett at damore.org
Fri Mar 25 10:20:20 PDT 2011


On Fri, 2011-03-25 at 18:15 +0100, Joerg Schilling wrote:
> Alan Coopersmith <alan.coopersmith at oracle.com> wrote:
> 
> > On 03/17/11 04:51 PM, Richard Lowe wrote:
> > > 63,000 lines of diff to go
> >
> > http://lucasr.org/2011/01/29/micro-commits/
> 
> Making commits the right way is something that is really important if you like 
> to make a project able to act as an upstream.
> 
> Illumos currently is not a nice upstream, importing commits from Illumos into 
> Schillix-ON is a hard job for most commits as the commits usually combine 
> unrelated tasks. If you like to import one of the sub-tasks but cannot import 
> another because it is incompatible, deltas need to be imported manually on a 
> file base.

There have been a few *pushes* with multiple commits, but those pushes
(largely from Rich Lowe) are well formed as separate commits with
separate bugs.

(Looking back, I realize I have a couple of commits that violate this
principle.  That's an artifact of me working on multiple things in the
same source tree, and not using something like mq or separate source
trees.  I'll endeavor to do a better job of breaking these apart in the
future.  Admittedly, I had not been so concerned about other potential
downstreams like Schillix-ON in my work here.)


> 
> In order to act nicely to potential downstrams, Schillix-ON tries to do it's 
> best to avoid bundling unrelated tasks.
> 
> One of the worst things in the ksh93 cluster commit is that there is a 
> Copyright date change in files that otherwise stay the same. This is something 
> I can try to ignore when reading the hg exort output, but it poisons the webrev.

There's a lot of unnecessary change in that webrev, agreed!

> I will probably need at least a week in order to do at least a basic review.

No rush.  There's enough substantive review feedback with actionable
items that the current ksh93 cluster won't be going in.

I've pretty much insisted that the ksh93 team break this commit up.
I've also offered that if they don't do this, then we will break it up
for them.

	- Garrett





More information about the Developer mailing list