🚀 go-pugleaf

I noticed recently that cc (without -ansi) and gcc on SGIs running
irix-4.0.5 have ANSI ctype macros that aren't ANSI.  The problem is
that all macros but isascii are undefined for non-ASCII character codes.
E.g. compiled with `cc' this program aborts

  #include <ctype.h>
  main() {if (isprint (0377)) abort(); exit(0);}

With `cc -ansi' it exits normally.
The only difference I could see is that plain `cc' uses /usr/lib/crt1.o
and `cc -ansi' uses /usr/lib/acrt1.o.

I want gcc to get the ANSI ctype tables, so I changed the specs file
so that gcc uses acrt1.o.

So far, this is the only difference I know of -- other than what nm
told me: that a couple of entrypoints like `environ' are in crt1.o, but
not in acrt1.o.

Is there any reason not to make gcc use the ANSI start file?

-- 
Jim Meyering             | Member of the League for Programming Freedom
meyering@cs.utexas.edu   | Contact me or lpf@uunet.uu.net for more info.