The new cross-build allows 9.1.1 to build master, so the pressure to
release is off, but perhaps you are still anxious to see it happen?

If so you might want to look at why `make check' fails on i386.  New
flonum tests reveal a snafu in SIGFPE handling that leaves a bogus
word in the heap.  The bogosity trips up find_block_address's heap
scan -- a "linkage section" where none should be.  I saved 8 or so
previous scan pointers and, at the point of the error, see valid
flonums and compiled-closures before and after the bogosity: off by
one?  My "fix" thus far is this kludge, which makes debugging ANY trap
a pain.

diff --git a/src/microcode/uxtrap.c b/src/microcode/uxtrap.c
index d1eb807..ce1bf28 100644
--- a/src/microcode/uxtrap.c
+++ b/src/microcode/uxtrap.c
@@ -48,6 +48,8 @@ USA.
 #  undef ENABLE_TRAP_RECOVERY
 #endif
 
+#undef ENABLE_TRAP_RECOVERY
+
 extern const char * find_signal_name (int);
 extern void UX_dump_core (void);
 extern void * initial_C_stack_pointer;

_______________________________________________
MIT-Scheme-devel mailing list
MIT-Scheme-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/mit-scheme-devel

Reply via email to