Re: The value of LilyPond, according to Ohloh

[David Kastrup]

Han-Wen Nienhuys <address@hidden> writes:

> On Mon, Nov 7, 2011 at 11:38 PM, David Kastrup <address@hidden> wrote:
>>> I think the number is way off.  AFAICS at
>>> http://www.ohloh.net/p/lilypond/enlistments?page=2, they counted all
>>> stable branches as separate.
>>>
>>> Last time I looked lilypond was around 100k lines of code (a quick
>>> look says it 200k now; I suspect the GNU headers as a cause). 390 man
>>> years looks exaggeratedl, since I wrote most of it (at least before
>>> 2007), and it took me much less than 300 years.
>>
>> Your output is likely not average for the industry.  Writing and
>> committing a half-hour change took you just half an hour instead of five
>> hours of red tape from you and three hours of red tape each from three
>> other persons.  We are getting there, but that's not what you have been
>> working with.
>
> In the early days (the "good old days") we didn't have a regression
> test (I checked in the first version in may 2006). Beyond all kinds of
> interesting stuff, I also committed several half hour changes that had
> me staring at gdb for hours (and sometimes days) on end.

I committed several changes recently that had me staring at bison output
tables for weeks.  Call me stupid, but argument scanning in the grammar
is specified from back to front while actually done front to back.
Which means that you need to invert time and causality when thinking
about the rules.  I think there was some jocular programming language
that had COMEFROM instead of GOTO.  Intercal or something?

Except that this is not a joke, and I don't think there is a better way.

-- 
David Kastrup