Module Name: src
Committed By: rillig
Date: Fri Apr 14 18:42:31 UTC 2023
Modified Files:
src/usr.bin/xlint/lint1: tree.c
Log Message:
lint: document suppressed LDBL_MAX warning more accurately
The crucial point is not that lint is cross-compiled but that tools/lint
is run on the usr.bin/xlint code, and in that situation, the platform
mismatch happens.
To generate a diff of this commit:
cvs rdiff -u -r1.512 -r1.513 src/usr.bin/xlint/lint1/tree.c
Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.
Modified files:
Index: src/usr.bin/xlint/lint1/tree.c
diff -u src/usr.bin/xlint/lint1/tree.c:1.512 src/usr.bin/xlint/lint1/tree.c:1.513
--- src/usr.bin/xlint/lint1/tree.c:1.512 Wed Apr 12 19:09:48 2023
+++ src/usr.bin/xlint/lint1/tree.c Fri Apr 14 18:42:31 2023
@@ -1,4 +1,4 @@
-/* $NetBSD: tree.c,v 1.512 2023/04/12 19:09:48 rillig Exp $ */
+/* $NetBSD: tree.c,v 1.513 2023/04/14 18:42:31 rillig Exp $ */
/*
* Copyright (c) 1994, 1995 Jochen Pohl
@@ -37,7 +37,7 @@
#include <sys/cdefs.h>
#if defined(__RCSID)
-__RCSID("$NetBSD: tree.c,v 1.512 2023/04/12 19:09:48 rillig Exp $");
+__RCSID("$NetBSD: tree.c,v 1.513 2023/04/14 18:42:31 rillig Exp $");
#endif
#include <float.h>
@@ -1508,18 +1508,21 @@ floating_error_value(tspec_t t, ldbl_t l
if (t == DOUBLE)
return lv < 0 ? -DBL_MAX : DBL_MAX;
/*
- * When lint is compiled on x86_64 to check for sparc64, it uses the
- * type 'long double' from x86_64, which is the Intel 80-bit format.
- * The constant LDBL_MAX comes from the sparc64 preprocessor though
- * and uses the IEEE-754-binary128 format, with the same exponent
- * range but a wider mantissa.
+ * When NetBSD is cross-built in MKLINT=yes mode on x86_64 for
+ * sparc64, tools/lint checks this code while building usr.bin/xlint.
+ * In that situation, lint uses the preprocessor for sparc64, in which
+ * the type 'long double' is IEEE-754-binary128, affecting the macro
+ * LDBL_MAX below. The type 'long double', as well as the strtold
+ * implementation, comes from the host platform x86_64 though, where
+ * 'long double' consumes 128 bits as well but only uses 80 of them.
+ * The exponent range of the two 'long double' types is the same, but
+ * the maximum finite value differs due to the extended precision on
+ * sparc64.
*
- * To properly handle this situation, lint would have to implement the
- * floating-point types in a platform-independent way, which is not
- * worth the effort, given how few programs practically use 'long
- * double'.
- *
- * This caveat only affects cross builds.
+ * To properly handle the data types of the target platform, lint
+ * would have to implement the floating-point types in a
+ * platform-independent way, which is not worth the effort, given how
+ * few programs practically use 'long double'.
*/
/* LINTED 248: floating-point constant out of range */
ldbl_t max = LDBL_MAX;
