svn 15195

[Rob Landley]

On Wednesday 31 May 2006 5:34 pm, Bernhard Fischer wrote:
> On Wed, May 31, 2006 at 05:14:56PM -0400, Rob Landley wrote:
> >Finally getting around to reviewing this:
>
> [snip]
> See ML for "struct option" (or the thread you started WRT
> "The getopt long thing breaks the tree").
> It's a non-standard extension. That knob installs a facility to
> circumvent that assertion fault.
>
> Put short:
> getopt_long breaks the build.

Except that getopt_long has been in the tree for years.  When you say "breaks 
the build" you mean "I'd like to modify busybox for a platform it's never 
supported before".

The second might be reasonable.  The first, by itself, is simply wrong.  The 
"standard" that busybox was developed under was Debian.  That was the 
expanded to generic Linux distributions.  Now you want to expand it off of 
Linux.  That's fine, but don't treat what it is now as unquestionably broken 
when what you want to do is something _new_.

> referencing a non-standard so called 'struct option' breaks the build.

The "utilities" section of susv3 is something we've been using as a guideline.  
This is actually fairly _recent_, and it tells us what functionality we 
should implement.

The operating system section of that document is totally flipping irrelevant 
to us.  We've never cared about it, and I'm not going to start.  BusyBox was 
designed against glibc and ported to uClibc early on.  That's _it_ in terms 
of what we supported until very recently.  You could get a subset of our 
functionality to work elsewhere, but this was by sheer coincidence.

Saying that stuff we use (like gcc extensions) is "nonstandard" is COMPLETELY 
IRRELEVANT and I DO NOT CARE.

Ok?

Saying "I need to change this in order to build on the native operating system 
of my Nintendo Gameboy" is at least a specific objection, and the only reason 
that so many strange devices came up in the first place was that Linux had 
first been ported to them.  BusyBox is Linux software, Linux runs on lots of 
strange hardware, I'm happy to make BusyBox run anywhere that Linux does.

Supporting other platforms is only IF THEY ARE EASY.  Porting BusyBox off of 
Linux is a side issue at best.  When non-Linux platforms come at the EXPENSE 
of working on Linux, I put my foot down and stop it.

> I assert that qualifies as plastic, yes? ;)

Go read The Innovator's Dilemma by Clayton Christensen if you want to 
understand my tagline.

Rob
-- 
Never bet against the cheap plastic solution.