Re: PATCH: libjava NetBSD configuration

[Bryce McKinlay <bryce at waitaki dot otago dot ac dot nz>]

Krister Walfridsson wrote:

>NetBSD does not implement MD_FALLBACK_FRAME_STATE_FOR, so we need to
>set DIVIDESPEC and CHECKREFSPEC. This removes 21 failures in the
>test suite.
>
>Bootstrapped and checked on i386-unknown-netbsdelf1.5ZC.
>
>   /Krister
>
>
>2002-04-10  Krister Walfridsson  <cato@df.lth.se>
>
>        * configure.host: (*-*-netbsd*) Set DIVIDESPEC and CHECKREFSPEC.
>

Actually for 3.1 we should really set these for all targets that both:

- arn't linux on alpha/PPC/IA64/x86
- arn't using SJLJ exceptions

This patch should do that. Could someone on FreeBSD or NetBSD verify 
that this does the right thing please?

regards

Bryce.


2002-04-11  Bryce McKinlay  <bryce@waitaki.otago.ac.nz>

	* configure.host: Set can_unwind_signal on hosts which support it.
	Don't set CHECKREFSPEC and DIVIDESPEC for FreeBSD.
	* configure.in: Set CHECKREFSPEC and DIVIDESPEC if not using SJLJ
	exceptions and can_unwind_signal isn't set.
	* configure: Rebuilt.

Index: configure.in
===================================================================
RCS file: /cvs/gcc/gcc/libjava/configure.in,v
retrieving revision 1.124
diff -u -r1.124 configure.in
--- configure.in	27 Mar 2002 16:41:34 -0000	1.124
+++ configure.in	11 Apr 2002 04:45:15 -0000
@@ -153,6 +153,14 @@
 AC_LANG_RESTORE
 AC_MSG_RESULT($ac_exception_model_name)
 
+# If we are non using SJLJ exceptions, and this host does not have support 
+# for unwinding from a signal handler, enable checked dereferences and divides.
+if test $can_unwind_signal = no && test $enable_sjlj_exceptions = no; then
+  CHECKREFSPEC=-fcheck-references
+  DIVIDESPEC=-fuse-divide-subroutine
+  EXCEPTIONSPEC=
+fi
+
 dnl See if the user wants to disable java.net.  This is the mildly
 dnl ugly way that we admit that target-side configuration sucks.
 AC_ARG_ENABLE(java-net,
Index: configure.host
===================================================================
RCS file: /cvs/gcc/gcc/libjava/configure.host,v
retrieving revision 1.30
diff -u -r1.30 configure.host
--- configure.host	5 Apr 2002 04:17:11 -0000	1.30
+++ configure.host	11 Apr 2002 04:45:15 -0000
@@ -26,6 +26,8 @@
 #   sysdeps_dir		Directory containing system-dependent headers
 #   slow_pthread_self   The synchronization code should try to avoid 
 #			pthread_self calls by caching thread IDs in a hashtable
+#   can_unwind_signal   Set to "yes" if the EH unwinder supports throwing
+#			from a signal handler.
 
 libgcj_flags=
 libgcj_cflags=
@@ -36,6 +38,7 @@
 enable_hash_synchronization_default=no
 sysdeps_dir=generic
 slow_pthread_self=
+can_unwind_signal=no
 
 case "${target_optspace}:${host}" in
   yes:*)
@@ -58,7 +61,6 @@
 
 DIVIDESPEC=-fuse-divide-subroutine
 EXCEPTIONSPEC=-fnon-call-exceptions
-CHECKREFSPEC=
 
 # This case statement supports per-CPU defaults.
 case "${host}" in
@@ -106,8 +108,6 @@
 	with_libffi_default=no
 	PROCESS=Ecos
 	FILE=Posix
-	CHECKREFSPEC=-fcheck-references
-	EXCEPTIONSPEC=
 	enable_java_net_default=no
 	enable_getenv_properties_default=no
 	enable_main_args_default=no
@@ -115,18 +115,21 @@
 esac
 
 # This case statement supports generic port properties and may refine
-# the above per-CPU defaults.  Note: If your OS does not implement
-# MD_FALLBACK_FRAME_STATE_FOR, then you may want to set CHECKREFSPEC,
-# DIVIDESPEC (to fix >20 test cases) and EXCEPTIONSPEC (to reduce EH
-# bloat only) here.
+# the above per-CPU defaults.  Note: If your OS implements
+# MD_FALLBACK_FRAME_STATE_FOR, then you want to set can_throw_from_signal
+# here.
 case "${host}" in
+  i[34567]86*-linux* | \
+  powerpc*-linux* | \
+  alpha*-linux* | \
+  ia64-*)
+  	can_unwind_signal=yes
+	;;
   *-*-darwin*)
 	enable_hash_synchronization_default=no
 	slow_pthread_self=
 	;;
   *-*-freebsd*)
-	DIVIDESPEC=-fuse-divide-subroutine
-	CHECKREFSPEC=-fcheck-references
 	slow_pthread_self=
 	;;
 esac