Patch 8.2.5004

[Bram Moolenaar]

Patch 8.2.5004
Problem: Right shift on negative number does not work as documented.
Solution: Use a uvarnumber_T type cast.
Files: runtime/doc/eval.txt, src/eval.c, src/vim9expr.c,
src/vim9execute.c, src/charset.c, src/testdir/test_expr.vim


*** ../vim-8.2.5003/runtime/doc/eval.txt 2022-05-22 19:13:45.496327034 +0100
--- runtime/doc/eval.txt 2022-05-22 20:14:04.001311133 +0100
***************
*** 1138,1146 ****
*E1282* *E1283*
The "<<" and ">>" operators can be used to perform bitwise left or right shift
of the left operand by the number of bits specified by the right operand. The
! operands must be positive numbers. The topmost bit (sign bit) is always
! cleared for ">>". If the right operand (shift amount) is more than the
! maximum number of bits in a number (|v:numbersize|) the result is zero.


expr6 and expr7 *expr6* *expr7* *E1036* *E1051*
--- 1138,1147 ----
*E1282* *E1283*
The "<<" and ">>" operators can be used to perform bitwise left or right shift
of the left operand by the number of bits specified by the right operand. The
! operands are used as positive numbers. When shifting right with ">>" the
! topmost bit (somtimes called the sign bit) is cleared. If the right operand
! (shift amount) is more than the maximum number of bits in a number
! (|v:numbersize|) the result is zero.


expr6 and expr7 *expr6* *expr7* *E1036* *E1051*
***************
*** 1417,1422 ****
--- 1418,1427 ----
Decimal, Hexadecimal (starting with 0x or 0X), Binary (starting with 0b or 0B)
and Octal (starting with 0, 0o or 0O).

+ Assuming 64 bit numbers are used (see |v:numbersize|) an unsigned number is
+ truncated to 0x7fffffffffffffff or 9223372036854775807. You can use -1 to get
+ 0xffffffffffffffff.
+
*floating-point-format*
Floating point numbers can be written in two forms:

*** ../vim-8.2.5003/src/eval.c 2022-05-22 19:13:45.496327034 +0100
--- src/eval.c 2022-05-22 19:20:46.928504419 +0100
***************
*** 3091,3102 ****
rettv->vval.v_number =
rettv->vval.v_number << var2.vval.v_number;
else
- {
rettv->vval.v_number =
! rettv->vval.v_number >> var2.vval.v_number;
! // clear the topmost sign bit
! rettv->vval.v_number &= ~((uvarnumber_T)1 << MAX_LSHIFT_BITS);
! }
}

clear_tv(&var2);
--- 3091,3098 ----
rettv->vval.v_number =
rettv->vval.v_number << var2.vval.v_number;
else
rettv->vval.v_number =
! (uvarnumber_T)rettv->vval.v_number >> var2.vval.v_number;
}

clear_tv(&var2);
*** ../vim-8.2.5003/src/vim9expr.c 2022-05-22 19:13:45.500327035 +0100
--- src/vim9expr.c 2022-05-22 19:25:26.640619364 +0100
***************
*** 2719,2729 ****
else if (type == EXPR_LSHIFT)
tv1->vval.v_number = tv1->vval.v_number << tv2->vval.v_number;
else
! {
! tv1->vval.v_number = tv1->vval.v_number >> tv2->vval.v_number;
! // clear the topmost sign bit
! tv1->vval.v_number &= ~((uvarnumber_T)1 << MAX_LSHIFT_BITS);
! }
clear_tv(tv2);
--ppconst->pp_used;
}
--- 2719,2726 ----
else if (type == EXPR_LSHIFT)
tv1->vval.v_number = tv1->vval.v_number << tv2->vval.v_number;
else
! tv1->vval.v_number =
! (uvarnumber_T)tv1->vval.v_number >> tv2->vval.v_number;
clear_tv(tv2);
--ppconst->pp_used;
}
*** ../vim-8.2.5003/src/vim9execute.c 2022-05-22 19:13:45.500327035 +0100
--- src/vim9execute.c 2022-05-22 19:21:57.620533615 +0100
***************
*** 4096,4107 ****
case EXPR_RSHIFT: if (arg2 > MAX_LSHIFT_BITS)
res = 0;
else
! {
! res = arg1 >> arg2;
! // clear the topmost sign bit
! res &= ~((uvarnumber_T)1
! << MAX_LSHIFT_BITS);
! }
break;
default: break;
}
--- 4096,4102 ----
case EXPR_RSHIFT: if (arg2 > MAX_LSHIFT_BITS)
res = 0;
else
! res = (uvarnumber_T)arg1 >> arg2;
break;
default: break;
}
*** ../vim-8.2.5003/src/charset.c 2022-05-21 20:16:51.003567195 +0100
--- src/charset.c 2022-05-22 19:57:32.935030889 +0100
***************
*** 2002,2007 ****
--- 2002,2008 ----
}
else
{
+ // prevent a larg unsigned number to become negative
if (un > VARNUM_MAX)
un = VARNUM_MAX;
*nptr = (varnumber_T)un;
*** ../vim-8.2.5003/src/testdir/test_expr.vim 2022-05-22 19:13:45.500327035 +0100
--- src/testdir/test_expr.vim 2022-05-22 19:48:17.396918194 +0100
***************
*** 958,963 ****
--- 958,965 ----
call assert_equal(0, 0 >> 4)
call assert_equal(0, 999999 >> 100)
call assert_equal(0, 999999 << 100)
+ call assert_equal(-1, -1 >> 0)
+ call assert_equal(-1, -1 << 0)
VAR a = 8
VAR b = 2
call assert_equal(2, a >> b)
***************
*** 976,981 ****
--- 978,992 ----
for i in range(0, v:numbersize - 2)
LET val = and(val, invert(1 << i))
endfor
+ #" -1 has all the bits set
+ call assert_equal(-2, -1 << 1)
+ call assert_equal(-4, -1 << 2)
+ call assert_equal(-8, -1 << 3)
+ if v:numbersize == 64
+ call assert_equal(0x7fffffffffffffff, -1 >> 1)
+ call assert_equal(0x3fffffffffffffff, -1 >> 2)
+ call assert_equal(0x1fffffffffffffff, -1 >> 3)
+ endif
call assert_equal(0, val)
#" multiple operators
call assert_equal(16, 1 << 2 << 2)
*** ../vim-8.2.5003/src/version.c 2022-05-22 19:13:45.500327035 +0100
--- src/version.c 2022-05-22 19:26:47.328652290 +0100
***************
*** 736,737 ****
--- 736,739 ----
{ /* Add new patch number below this line */
+ /**/
+ 5004,
/**/

--
hundred-and-one symptoms of being an internet addict:
267. You get an extra phone line so you can get phone calls.

/// Bram Moolenaar -- Br...@Moolenaar.net -- http://www.Moolenaar.net \\\
/// \\\
\\\ sponsor Vim, vote for features -- http://www.Vim.org/sponsor/ ///
\\\ help me help AIDS victims -- http://ICCF-Holland.org ///