possible fix for valgrinding problem on longer running kernels (2.4.18-SuSE, update 17 days)
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@506 a5019735-40e9-0310-863c-91ae7b9d1cf9
diff --git a/coregrind/vg_include.h b/coregrind/vg_include.h
index 40506c1..31ba8e9 100644
--- a/coregrind/vg_include.h
+++ b/coregrind/vg_include.h
@@ -788,10 +788,12 @@
the initial stack, which we can't move, is allocated here.
VG_(scheduler_init) checks this. Andrea Archelangi's 2.4 kernels
have been rumoured to start stacks at 0x80000000, so that too is
- considered.
+ considered. It seems systems with longer uptimes tend to to use
+ stacks which start at 0x40000000 sometimes.
*/
#define VG_STARTUP_STACK_BASE_1 (Addr)0xC0000000
#define VG_STARTUP_STACK_BASE_2 (Addr)0x80000000
+#define VG_STARTUP_STACK_BASE_3 (Addr)0x40000000
#define VG_STARTUP_STACK_SMALLERTHAN 0x100000 /* 1024k */
#define VG_STACK_MATCHES_BASE(zzstack, zzbase) \