[m-rev.] for review: simplify profiler feedback code

[Paul Bone]

On Sun, Nov 30, 2014 at 01:42:28AM +1100, Zoltan Somogyi wrote:
> On Sat, 29 Nov 2014 19:31:24 +1100 (EST), "Zoltan Somogyi" <zoltan.somogyi at runbox.com> wrote:
> > Ah. It seems that we were talking past each other. The comment on
> > read_or_create in mdbcomp/feedback.m said "ProgramName is the name
> > of the program that generated this feedback file". I took that to mean
> > that it is the name of mdprof_create_feedback, or a similar Mercury tool
> > that creates feedback files. Your comment makes it clear to me that it is
> > meant to be the name NOT of the program that generated the feedback
> > file, but of the program the feedback file was generated FOR, i.e. the
> > name of the program whose execution created the Deep.data file
> > the feedback was created FROM. That makes much more sense.
> > I will fix the misleading comment, and I will look over all the code
> > that handles these program names with that in mind.
> 
> I have now done that, and modified all the code touching that field
> according to this better interpretation. I also addressed all the issues
> identified in the review of the original diff. The result is the updated
> diff attached to this mail. Since the diff for mdbcomp/feedback.m
> is quite confusing due to the big changes to that file, I have attached
> it separately as well.
> 
> I have bootchecked this in a deep profiling grade, and the only failures
> were test cases that always fail in such grades. (I will fix these separately.)
> 
> Paul, do you want to review the updated diff, and if yes, then are you ok
> with a post-commit review?
> 

I've taken a quick look now (I didn't have any major concerns anyway).  Can
you check the spelling of Jerome's name, It looks like you've spelt it
Jereon.  I that the correct spelling is Jerömé, but I understand how you
feel about non-ASCII characters.

Also I think that pointing users at the date in a git repository can be a
bit strange, as patches may not always be applied in date-order.  Instead I
suggest poƣnting users at the 14.01.1 release, in this case the feedback
code in that release and in the master branch will be the same.

I think there's also more of Jerömé's code that can be removed.  I can do
this post-commit.

Cheers.



-- 
Paul Bone