[comp.lang.c++] RETBUG flag in building cfront version 1.2

peter@mit-amt.MEDIA.MIT.EDU (Peter Schroeder) (09/09/89)

Does anybody know whether I can undefine RETBUG when building cfront 1.2
for the HP9000/835?

I tried some tests by hand and found that this _plain_silly business costs
a lot of time ( apparently the HP compiler optimizer pass does not see that
the whole thing is bogus ) when running on an HP9000/835. However I am
somewhat scared just turning that flag off, since I don't understand this
bug that the RETBUG flag is a workaround for.

Also has anybody using cfront 1.2 on the HP9000/835 gotten the form()
library call to work properly? Apparently its assumptions about stack
layout don't hold on the HP9000/835 so that things like

double a = 1.2345;
cout << form( "%F", a ) << "\n";

just print garbage.

Any suggestions/help is much appreciated!

Peter

peter@media-lab.media.mit.edu

shankar@hpclscu.HP.COM (Shankar Unni) (09/11/89)

> Does anybody know whether I can undefine RETBUG when building cfront 1.2
> for the HP9000/835?

Go ahead. I can't think of what that means. I looked at the sources and
there are some nice circular pointers there (print.c says look at simpl.c,
and vice versa).

If you've got a recent system (HP-UX 3.0 or greater) on the 835, I'd say go
ahead and nuke that RETBUG. (Of course, if you then find some problem,
*please* let us know).

BTW, the RETBUG stuff has been totally nuked from the 2.0 source, so I
presume AT&T decided it wasn't necessary.

> I tried some tests by hand and found that this _plain_silly business costs
> a lot of time ( apparently the HP compiler optimizer pass does not see that
> the whole thing is bogus ) when running on an HP9000/835. However I am

Oh, I'm sure.

> Also has anybody using cfront 1.2 on the HP9000/835 gotten the form()
> library call to work properly? Apparently its assumptions about stack
> layout don't hold on the HP9000/835 so that things like

Aha! The old "form()" question.

For 2.0, go to streamdefs.h and uncomment the define for
  #define VSPRINTF vsprintf
and watch it work. For 1.2, I'll mail you a source patch.
-----
Shankar Unni                                   E-Mail: 
Hewlett-Packard California Language Lab.     Internet: shankar@hpda.hp.com
Phone : (408) 447-5797                           UUCP: ...!hplabs!hpda!shankar