summaryrefslogtreecommitdiffstats
path: root/loc2c.c
diff options
context:
space:
mode:
authorFrank Ch. Eigler <fche@elastic.org>2009-11-24 15:46:41 -0500
committerFrank Ch. Eigler <fche@elastic.org>2009-12-20 12:17:07 -0500
commita5a674fa2c2447e9dd34c8c8f7b1e5756a32df4d (patch)
tree9a5a7723d9552f302f4958b66079b545a92efc87 /loc2c.c
parent39b375aeafc43e00ecf68d0151e7df0597139949 (diff)
downloadsystemtap-steved-a5a674fa2c2447e9dd34c8c8f7b1e5756a32df4d.tar.gz
systemtap-steved-a5a674fa2c2447e9dd34c8c8f7b1e5756a32df4d.tar.xz
systemtap-steved-a5a674fa2c2447e9dd34c8c8f7b1e5756a32df4d.zip
PR10601: comment on loc2c's use of [u]intptr_t for temp values
Diffstat (limited to 'loc2c.c')
-rw-r--r--loc2c.c7
1 files changed, 7 insertions, 0 deletions
diff --git a/loc2c.c b/loc2c.c
index d3ec55fa..a7473784 100644
--- a/loc2c.c
+++ b/loc2c.c
@@ -30,6 +30,13 @@
#define N_(x) x
+/* NB: PR10601 may make one suspect that intptr_t and uintptr_t aren't
+ right, for example on a 64-bit kernel targeting a 32-bit userspace
+ process. At least these types are always at least as wide as
+ userspace (since 64-bit userspace doesn't run on a 32-bit kernel).
+ So as long as deref() and {fetch,store}_register() widen/narrow
+ their underlying values to these, there should be no problem. */
+
#define STACK_TYPE "intptr_t" /* Must be the signed type. */
#define UTYPE "uintptr_t" /* Must be the unsigned type. */
#define SFORMAT "%" PRId64 "L"