[src/trunk]: src/tests/usr.bin/xlint/lint1 tests/lint: document why in ic_exp...

[rillig <rillig%NetBSD.org@localhost>]

details:   https://anonhg.NetBSD.org/src/rev/3f737c49b7ec
branches:  trunk
changeset: 368368:3f737c49b7ec
user:      rillig <rillig%NetBSD.org@localhost>
date:      Thu Jul 07 18:11:29 2022 +0000

description:
tests/lint: document why in ic_expr, '&' does not need before_conversion

diffstat:

 tests/usr.bin/xlint/lint1/msg_132.c |  19 ++++++++++++++++++-
 1 files changed, 18 insertions(+), 1 deletions(-)

diffs (30 lines):

diff -r 1851b09e2d1c -r 3f737c49b7ec tests/usr.bin/xlint/lint1/msg_132.c
--- a/tests/usr.bin/xlint/lint1/msg_132.c       Thu Jul 07 13:27:02 2022 +0000
+++ b/tests/usr.bin/xlint/lint1/msg_132.c       Thu Jul 07 18:11:29 2022 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: msg_132.c,v 1.23 2022/07/06 22:26:31 rillig Exp $      */
+/*     $NetBSD: msg_132.c,v 1.24 2022/07/07 18:11:29 rillig Exp $      */
 # 3 "msg_132.c"
 
 // Test for message: conversion from '%s' to '%s' may lose accuracy [132]
@@ -329,3 +329,20 @@
        /* expect+1: warning: conversion from 'long long' to 'signed char' may lose accuracy [132] */
        s8 = s64 % 1;
 }
+
+void
+test_ic_bitand(void)
+{
+       /*
+        * ic_bitand assumes that integers are represented in 2's complement,
+        * and that the sign bit of signed integers behaves like a value bit.
+        * That way, the following expressions get their constraints computed
+        * correctly, regardless of whether ic_expr takes care of integer
+        * promotions or not.  Compare ic_mod, which ignores signed types.
+        */
+
+       u8 = u8 & u16;
+
+       /* expect+1: warning: conversion from 'unsigned int' to 'unsigned char' may lose accuracy [132] */
+       u8 = u16 & u32;
+}