Re: build system work

[Graham Percival]

On Fri, Mar 04, 2011 at 11:38:19AM -0700, Carl Sorensen wrote:
> On 3/4/11 11:09 AM, "Colin Campbell" <address@hidden> wrote:
> 
> > On 11-03-04 10:54 AM, Carl Sorensen wrote:
> >> On 3/4/11 10:31 AM, "Colin Campbell"<address@hidden>  wrote:
> >> 

[subject: wiki or not?]

> >>> I'm the noob here, so I'll defer to greater wisdom.

Ok, for the record: we're still gathering opinions about wiki vs.
CG.  The most important opinions come from the people offering to
actually do the work, of course.  So far, that means Colin and
Phil.

> >> My preference would be to have an issue on google code, with a priority of
> >> low, under which one-time users could post their experiences.

Please no.  We'll be gathering approximately 15 pdf-pages of docs;
having that amount of info as a series of comments on google code
wouldn't work.

(unless I'm misunderstanding this comment...?)

> >> I'd prefer *not* to put patches for the CG on Rietveld; I'd rather just see
> >> them pushed.  I think there's no sense discussing them at this point.
> >> Better just to capture them.

Yes, absolutely.

> >> The main difficulty I see with Graham's original proposed workflow is that
> >> not everybody has git push access, so we need somebody who will volunteer 
> >> to
> >> push patches from those without access.

I was going to dump them all on James.  It'll be a good learning
experience for him, and as long as all patches only touch the file
Documentation/contributor/build-notes.itexi , nobody else needs to
care if anything goes wrong.  (as long as the docs can still
compile, at least).

> The proposal was to start by documenting what the build system
> does.  And he estimated that the documentation process would
> take 50 hours of development time.

Yes.

> But I don't think that project needs any visibility on frogs --
> it's way over the head of frogs.  Anybody who is contributing to
> build system revision is clearly a developer, IMO.

Yes and no.  The only two volunteers so far clearly *are* Frogs.


My other big qualm is the amount of work I'll be putting in, and
the resulting inefficiencies.  For better or worse, I think that I
know more about the doc+web build process than anybody else.  I
would not be surprised if I could describe (+ learn if necessary)
the build process 10 times faster than Colin and/or Phil.  I'm not
trying to brag here about me being fast at learning -- it's not
that at all!  This is simply because I spent most of Fall 2009
spending 40+ hours a week working on lilypond, most of which was
on the build process.  I've already spent dozens of hours puzzling
through bits and pieces of the build system.

However, I am now absolutely committed to 10 hours a week for the
duration of my PhD.  And almost all that time is already accounted
for.  This means that it is extremely likely that Phil and/or
Colin will spent 5+ hours trying to figure out something that I
could explain in a 10-minute email.

I can't think of a good solution, though.  It sucks, but that's
life?  I refuse to budge on my allocated time; my PhD has already
suffered enough from my LilyPond work.  Other than suggesting that
"if something seems really puzzling, stop and ask Graham a quick
question", I can't think of a way to alleviate this problem.

Oh, there _is_ one way: if more people took on more of my current
duties, I'd have more time left in my 10 hours, and could thus
spend more time helping this process.  The biggest/easiest time
sink for me right now is checking and bestowing Patch-review
labels on issues:
http://lilypond.org/doc/v2.13/Documentation/contributor/patch-handling
If anybody thinks they can handle that -- all you need is to
compile the source! -- then I could be a bit more flexible with my
time.

Cheers,
- Graham