/* $NetBSD: expr_fold.c,v 1.15 2024/03/10 19:45:14 rillig Exp $ */ # 3 "expr_fold.c" /* * Test folding of constant expressions. */ /* lint1-extra-flags: -h -X 351 */ /* * On ILP32 platforms, the integer constant 2147483648 cannot be represented * as 'int' or 'long', therefore it becomes 'long long'. This would * influence the type names in the diagnostics. */ /* lint1-only-if: lp64 */ void take_bool(_Bool); void take_int(int); void take_uint(unsigned int); /* * C99 6.4.4.1p5 defines that decimal integer constants without suffix get * one of the signed integer types. On the other hand, octal and hexadecimal * constants get either a signed or unsigned type, whichever fits first. */ void fold_uplus(void) { take_int(+(0)); take_int(+(2147483647)); /* expect+1: warning: conversion of 'long' to 'int' is out of range, arg #1 [295] */ take_int(+(2147483648)); /* expect+1: warning: conversion of 'long' to 'int' is out of range, arg #1 [295] */ take_int(+(4294967295)); take_uint(+(0)); take_uint(+(2147483647)); take_uint(+(2147483648)); take_uint(+(4294967295)); /* * Hexadecimal constants and constants with the suffix 'U' get either * a signed or an unsigned integer type, so no warning here. */ take_uint(+(2147483648U)); take_uint(+(0x80000000)); take_uint(+(0x80000000U)); } void fold_uminus(void) { take_int(-(0)); take_int(-(2147483647)); take_int(-(2147483648)); /* The '-' is an operator, it is not part of the integer constant. */ take_int(-2147483648); /* expect+1: warning: '2147483647 + 1' overflows 'int' [141] */ take_int(-(2147483647 + 1)); /* expect+1: warning: '-(-2147483648)' overflows 'int' [141] */ take_int(-(-2147483647 - 1)); /* expect+1: warning: conversion of 'long' to 'int' is out of range, arg #1 [295] */ take_int(-(4294967295)); take_uint(-(0)); /* expect+1: warning: conversion of negative constant to unsigned type, arg #1 [296] */ take_uint(-(2147483647)); /* expect+1: warning: conversion of negative constant to unsigned type, arg #1 [296] */ take_uint(-(2147483648)); /* expect+1: warning: conversion of negative constant to unsigned type, arg #1 [296] */ take_uint(-(4294967295)); } void fold_compl(void) { take_int(~(0)); take_int(~(2147483647)); /* expect+1: warning: conversion of 'long' to 'int' is out of range, arg #1 [295] */ take_int(~(2147483648)); /* expect+1: warning: conversion of 'long' to 'int' is out of range, arg #1 [295] */ take_int(~(4294967295)); /* expect+1: warning: conversion of negative constant to unsigned type, arg #1 [296] */ take_uint(~(0)); /* expect+1: warning: conversion of negative constant to unsigned type, arg #1 [296] */ take_uint(~(2147483647)); /* expect+1: warning: conversion of negative constant to unsigned type, arg #1 [296] */ take_uint(~(2147483648)); /* expect+1: warning: conversion of negative constant to unsigned type, arg #1 [296] */ take_uint(~(4294967295)); } void fold_mult(void) { take_int(32767 * 65536); /* expect+1: warning: '32768 * 65536' overflows 'int' [141] */ take_int(32768 * 65536); /* expect+1: warning: '65536 * 65536' overflows 'int' [141] */ take_int(65536 * 65536); take_uint(32767 * 65536U); take_uint(32768 * 65536U); /* expect+1: warning: '65536 * 65536' overflows 'unsigned int' [141] */ take_uint(65536 * 65536U); } void fold_div(void) { /* expect+1: error: division by 0 [139] */ take_int(0 / 0); /* expect+1: warning: conversion of 'long' to 'int' is out of range, arg #1 [295] */ take_int(-2147483648 / -1); } void fold_mod(void) { /* expect+1: error: modulus by 0 [140] */ take_int(0 % 0); /* expect+1: error: modulus by 0 [140] */ take_int(0 % 0U); /* expect+1: error: modulus by 0 [140] */ take_int(0U % 0); /* expect+1: error: modulus by 0 [140] */ take_int(0U % 0U); take_int(-2147483648 % -1); } void fold_plus(void) { /* expect+1: warning: '2147483647 + 1' overflows 'int' [141] */ take_int(2147483647 + 1); /* Assume two's complement, so no overflow. */ take_int(-2147483647 + -1); /* expect+1: warning: '-2147483647 + -2' overflows 'int' [141] */ take_int(-2147483647 + -2); /* * No overflow since one of the operands is unsigned, therefore the * other operand is converted to unsigned as well. * See C99 6.3.1.8p1, paragraph 8 of 10. */ /* wrong integer overflow warning before tree.c 1.338 from 2021-08-19 */ take_uint(2147483647 + 1U); /* wrong integer overflow warning before tree.c 1.338 from 2021-08-19 */ take_uint(2147483647U + 1); } void fold_minus(void) { /* expect+1: warning: '2147483647 - -1' overflows 'int' [141] */ take_int(2147483647 - -1); /* Assume two's complement. */ take_int(-2147483647 - 1); /* expect+1: warning: '-2147483647 - 2' overflows 'int' [141] */ take_int(-2147483647 - 2); take_int(0 - 2147483648); /* expect+1: warning: '0 - 2147483648' overflows 'unsigned int' [141] */ take_uint(0 - 2147483648U); } void fold_shl(void) { /* expect+1: warning: '16777216 << 24' overflows 'int' [141] */ take_int(1 << 24 << 24); /* expect+1: warning: '16777216 << 24' overflows 'unsigned int' [141] */ take_uint(1U << 24 << 24); /* expect+1: warning: shift amount 104 is greater than bit-size 32 of 'unsigned int' [122] */ take_uint(1U << 24 << 104); } void fold_shr(void) { take_int(16777216 >> 24); take_int(16777216 >> 25); /* expect+1: warning: shift amount 104 is greater than bit-size 32 of 'int' [122] */ take_int(16777216 >> 104); } void fold_lt(void) { take_bool(1 < 3); take_bool(3 < 1); } void fold_le(void) { take_bool(1 <= 3); take_bool(3 <= 1); } void fold_ge(void) { take_bool(1 >= 3); take_bool(3 >= 1); } void fold_gt(void) { take_bool(1 > 3); take_bool(3 > 1); } void fold_eq(void) { take_bool(1 == 3); take_bool(3 == 1); } void fold_ne(void) { take_bool(1 != 3); take_bool(3 != 1); } void fold_bitand(void) { take_bool(1 & 3); take_bool(3 & 1); } void fold_bitxor(void) { take_bool(1 ^ 3); take_bool(3 ^ 1); } void fold_bitor(void) { take_bool(1 | 3); take_bool(3 | 1); } /* * The following expression originated in vndcompress.c 1.29 from 2017-07-29, * where line 310 contained a seemingly harmless compile-time assertion that * expanded to a real monster expression. * * __CTASSERT(MUL_OK(uint64_t, MAX_N_BLOCKS, MAX_BLOCKSIZE)); * * Before tree.c 1.345 from 2021-08-22, lint wrongly assumed that the result * of all binary operators were the common arithmetic type, but that was * wrong for the comparison operators. The expression '1ULL < 2ULL' does not * have type 'unsigned long long' but 'int' in default mode, or '_Bool' in * strict bool mode. */ struct ctassert5_struct { unsigned int member: /*CONSTCOND*/ 0xfffffffeU <= ((1ULL << 63) + 1 < 1 ? ~(1ULL << 63) : ~0ULL) / 0xfffffe00U ? 1 : -1; }; /* * Since Makefile.inc 1.21 from 2022-04-08 (which added -ftrapv) and before * tree.c 1.436 from 2022-04-20, lint crashed with an integer overflow when * calculating '-(uint64_t)INT64_MIN' in val_t.u.integer. */ void unary_minus_overflow(unsigned long long val) { if (val > -(unsigned long long)(-0x7fffffffffffffffL - 1)) return; }