Re: [Gpsbabel-code] Patch to geo.c

[Robert L. <rob...@us...>]

Gustavo Niemeyer wrote:

> > (And like most non-trivial projects, we have some nice code and some
> > icky code. :-)
>
> In that case, I was lukcy to touch a nice part. :-)

geo.c got refactored this spring to use my "generic xml" layer which
wraps expat into something rather more practical.  So 'geo' isn't nearly
as tortured as some of that code that's overdue for a similar overhaul.

> I can even understand that they didn't want to bufferize a very
> large amount of data before passing to the callback. 

So instead they leave it to every caller to do it.  Of course, if you
don't do your own buffering, the code basically appears to work but
fails in rare data-dependent ways instead of failing a manner designed
to draw your attention to it.

Sorry, that's just not a good design for a library.

> I was beaten by that issue on another software as well.

The prosecution rests. :-)

RJL