Hi,
Rob Browning [EMAIL PROTECTED] writes:
Please test.
I only ran `make check' and it succeeds on `powerpc-unknown-linux-gnu'.
Thanks,
Ludovic.
___
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel
One more bit of info. The only change needed to get the build
to run to completion is to set SCM_DEBUG_CELL_ACCESSES to 1.
The resultant guile also runs without stack overflow.
___
Guile-devel mailing list
Guile-devel@gnu.org
Or even simpler, just change the SCM_VALIDATE_CELL definition in
gc.h to:
#define SCM_VALIDATE_CELL(cell, expr) (scm_remember_upto_here(cell), (expr))
This is kinda fun!
___
Guile-devel mailing list
Guile-devel@gnu.org
Actually the only cases that need the check are SCM_CELL_OBJECT_1 and
SCM_SET_CELL_OBJECT_0. But perhaps there's some way to embed that
asm business (in gc.h) into the SCM_VALIDATE_CELL macro.
I haven't used asm before, but here's a suggestion that seems to work:
#define
Bill Schottstaedt [EMAIL PROTECTED] writes:
#define SCM_VALIDATE_CELL(cell, expr) ({ __asm__ __volatile__ ( : : g
(cell)); expr; })
The asm version of scm_remember_upto_here_1 could be made into an
expression like that ({...}), as opposed to the current statement
style, if that helped make
scm_make_port_type returns scm_t_bits, but the various setting
functions like scm_set_port_mark take a long. Should they be
scm_t_bits too?
___
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel