From: Mathieu Bouchard <matju@...>
Date: 2001-01-08T07:27:21+09:00
Subject: [ruby-talk:8791] Re: 1.6.2 on Dec Alpha

> > > And on a Dec Alpha system with gcc -v giving:
> > > Reading specs from
> > > /usr/local/gnu/lib/gcc-lib/alphaev6-dec-osf4.0f/2.95.2/specs
> > > gcc version 2.95.2 19991024 (release)
> > > gcc -g -O2 -ansi  main.o dmyext.o libruby.a -lm   -o miniruby
> > > ./mkconfig.rb:44: stack level too deep (SystemStackError)
> > Got the same on a NetWinder (StrongARM 233 MHz) yesterday. It looks like
> > it stays stack too deep even though it's at depth 2. Trying:
> My "bodge" for this problem was to change the makefile so that -ansi
> was not used in the CFLAGS.  What I should do is this:

Here it is:
gcc -v
Reading specs from /usr/lib/gcc-lib/armv4l-unknown-linux-gnu/2.95.1/specs
gcc version 2.95.1 19990816 (release)

and dig this: there is no -ansi during compilation, but the stack problem
is there. 

I patched eval.c so that it gives me some info about the involved values:

            sprintf(foo,
                "too deep: stack_length(0)=%d, STACK_LEVEL_MAX=%d",
                stack_length(0),STACK_LEVEL_MAX);

./mkconfig.rb:44: too deep:
	stack_length(0)=157736579,
	STACK_LEVEL_MAX=1835008 (SystemStackError)

this seems like little too high a value.

gdb miniruby
run -Cext extmk.rb

print rb_gc_stack_start               #=> (VALUE *) 0xe24cb004

added a printf for __builtin_frame_address(0)

				      #=> 0xbfffb998

for some reason, in gc.c, in Init_stack(),
__builtin_frame_address(2) gives a value like 0xe24cb004 (???)

well that's all I'm trying for now. I know very little about this
architecture and stack issues.

matju