[APBeta] Interesting find

[Jeff Young]

Paul --

Even C++ compilers sometimes have a hard time with long, complex statements, and it does tend to vary from compiler version to version.

That being said, the C++ compilers almost always fail at compile time (rather than at run-time), and I've never seen one produce results that failed sometimes and not other times.

I did once find a chip bug in the 68020 (the early version which was shipped with some of the first NeXT systems).  It had something to do with a stack overflow when executing a certain floating point operation.  We just wrote the code slightly differently and Motorolla corrected the chip in the next batch.  But again, it was reproducible....

Cheers,
-- Jeff.

> -----Original Message-----
> From: apbeta-bounces at lists.astroplanner.net [mailto:apbeta-bounces at lists.astroplanner.net] On Behalf Of
> Paul Rodman
> Sent: Saturday, November 29, 2008 10:28 PM
> To: AstroPlanner Beta Testers
> Subject: [APBeta] Interesting find
>
> On Friday I spent most of the day trying to figure out why the code to
> draw Saturn in the FOV was causing reproducible hard crashes on
> Windows (fine on Mac OS). Eventually by simplifying the (rather
> complex math) code and spreading it out over a number of separate
> routines, I was able to get it to work.
>
> This is rather disturbing. It would appear to be either a compiler bug
> with complex math expressions or some kind of undetected stack
> overflow problem. I'm not sure what.
>
> The app is still unstable (i.e. it "randomly" hard crashes), and now
> I'm wondering if it has to do with other math code that gets executed
> occasionally (e.g. moon/sun position, etc.). This is turning me
> prematurely grey. I'll have a look at other bits of "nasty" code this
> week and see if I can do the same with them, in the hope that it will
> resolve the problems.
>
> Note: this code is virtually unchanged from that in V1.6, which makes
> me suspect the compiler.
>
> Paul R.
>
> _______________________________________________
> APBeta mailing list
> APBeta at lists.astroplanner.net
> http://lists.astroplanner.net/listinfo.cgi/apbeta-astroplanner.net