> I'm trying really hard not to punt here.  The egcs trunk is in such poor
> shape wrt stability that it's not even a remote possibility.  And now, alpha
> actually has a working compiler built off of 1.1-branch.  *grr!*

I thought the whole point of EGCS was to provide a place for smushing
together all the things people have hacked into GCC but never gotten
past Kenner and/or Stallman. So naturally it would be a mess for a while.

Just think of all those poor RedHat users who are unwittingly beta testing
egcs-1.0.2something ...

> Sometimes I wonder just how they sell service agreements and build GNUPro
> off of that beast.  I'm still hoping for better, though.

Having spent 5 years at Green Hills, I say to you: it can be done.

If a commercial compiler's release branch isn't on the order of a year
old, expect bugs. If it's older than a year, expect bugs, but not as many.
Still, you might be surprised at how usable a buggy compiler can be.

It used to shock me just how much you can ignore and still squeak out a
product that gets past QA; later I realized that a tool is not judged on
how many bugs it _has_ but rather on how many of those bugs actually keep
people from doing their jobs.

Of course, when all you do is build a HUGE tree of which 90% hasn't changed
in the last decade, and perhaps 30% of which is actually executed immediately
after building it, you're considerably more likely to notice new bugs, and be
unable to ignore them for very long.

Todd Whitesel
toddpw @ best.com