Source-Changes-HG archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

[src/trunk]: src/tests/usr.bin/xlint/lint1 tests/lint: add tests for warning ...



details:   https://anonhg.NetBSD.org/src/rev/612d7948535a
branches:  trunk
changeset: 960684:612d7948535a
user:      rillig <rillig%NetBSD.org@localhost>
date:      Fri Mar 26 16:19:43 2021 +0000

description:
tests/lint: add tests for warning about cast to character types

diffstat:

 tests/usr.bin/xlint/lint1/msg_247.c   |  47 ++++++++++++++++++++++++++++++++++-
 tests/usr.bin/xlint/lint1/msg_247.exp |   3 ++
 2 files changed, 49 insertions(+), 1 deletions(-)

diffs (66 lines):

diff -r 229b9b5097c2 -r 612d7948535a tests/usr.bin/xlint/lint1/msg_247.c
--- a/tests/usr.bin/xlint/lint1/msg_247.c       Fri Mar 26 16:05:19 2021 +0000
+++ b/tests/usr.bin/xlint/lint1/msg_247.c       Fri Mar 26 16:19:43 2021 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: msg_247.c,v 1.5 2021/03/14 22:24:24 rillig Exp $       */
+/*     $NetBSD: msg_247.c,v 1.6 2021/03/26 16:19:43 rillig Exp $       */
 # 3 "msg_247.c"
 
 // Test for message: pointer cast from '%s' to '%s' may be troublesome [247]
@@ -30,3 +30,48 @@
         */
        display = (PDisplay)arg;        /* expect: 247 */
 }
+
+/*
+ * C code with a long history that has existed in pre-C90 times already often
+ * uses 'pointer to char' where modern code would use 'pointer to void'.
+ * Since 'char' is the most general underlying type, there is nothing wrong
+ * with casting to it.  An example for this type of code is X11.
+ *
+ * Casting to 'pointer to char' may also be used by programmers who don't know
+ * about endianness, but that's not something lint can do anything about.  The
+ * code for these two use cases looks exactly the same, so lint errs on the
+ * side of fewer false positive warnings here. (after fixing the FIXME below)
+ */
+char *
+cast_to_char_pointer(struct Other *arg)
+{
+       return (char *)arg;             /* expect: 247 *//* FIXME */
+}
+
+/*
+ * In traditional C there was 'unsigned char' as well, so the same reasoning
+ * as for plain 'char' applies here.
+ */
+unsigned char *
+cast_to_unsigned_char_pointer(struct Other *arg)
+{
+       return (unsigned char *)arg;    /* expect: 247 *//* FIXME */
+}
+
+/*
+ * Traditional C does not have the type specifier 'signed', which means that
+ * this type cannot be used by old code.  Therefore warn about this.  All code
+ * that triggers this warning should do the intermediate cast via 'void
+ * pointer'.
+ */
+signed char *
+cast_to_signed_char_pointer(struct Other *arg)
+{
+       return (signed char *)arg;      /* expect: 247 */
+}
+
+char *
+cast_to_void_pointer_then_to_char_pointer(struct Other *arg)
+{
+       return (char *)(void *)arg;
+}
diff -r 229b9b5097c2 -r 612d7948535a tests/usr.bin/xlint/lint1/msg_247.exp
--- a/tests/usr.bin/xlint/lint1/msg_247.exp     Fri Mar 26 16:05:19 2021 +0000
+++ b/tests/usr.bin/xlint/lint1/msg_247.exp     Fri Mar 26 16:19:43 2021 +0000
@@ -1,1 +1,4 @@
 msg_247.c(31): warning: pointer cast from 'pointer to struct Other' to 'pointer to struct <unnamed>' may be troublesome [247]
+msg_247.c(48): warning: pointer cast from 'pointer to struct Other' to 'pointer to char' may be troublesome [247]
+msg_247.c(58): warning: pointer cast from 'pointer to struct Other' to 'pointer to unsigned char' may be troublesome [247]
+msg_247.c(70): warning: pointer cast from 'pointer to struct Other' to 'pointer to signed char' may be troublesome [247]



Home | Main Index | Thread Index | Old Index