blob: d95ed559238f9ac67bcd9225fa4814b73410ffea (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
|
* (Further) refactoring (s)printf formatting code:
[1] Remove direct calls to format_tree() in number to string (force_string)
routines. Fix CONVFMT = "%s"/"%c" crash; present in all known gawk versions.
Cache parsed format code in fmt_idx() ... DONE
[2] Try to format NaN/inf in main gawk code, and not in the individual handlers.
* In non-number related routines (substr, index, ..), fetch an integer directly
using get_number_si/get_number_ui to get the correct size at the int boundaries.
Use isinteger() to test for a floating-point value instead of get_number_d() and ...
May require additional tests e.g. SIZE_MAX <= UINT_MAX. In a nutshell, don't do this:
number => double => integer
* Restore constant-folding code for numbers.
* Consider special handling of integer-indexed arrays with non-double
number handlers.
- Choose a reasonable range for integer indices.
#if SIZEOF_LONG <= 8
typedef gawk_int_t long
#define GAWK_INT_MAX LONG_MAX
#define GAWK_INT_MIN LONG_MIN
#define GAWK_INT_BIT SIZEOF_LONG * 8 ### may not need this
#else
typedef gawk_int_t int
#define GAWK_INT_MAX INT_MAX
#define GAWK_INT_MIN INT_MIN
#define GAWK_INT_BIT SIZEOF_INT * 8
#endif
Install gaurd code in array handlers (or adjust the defs) for size
anything but 4 or 8 (overthinking?). Use gawk_int_t instead
of int32_t or int64_t.
- Use additional struct field number_fits_gawk_int()
(MPFR has similarly named routines, e.g. mpfr_fits_slong()),
and avoid checking after conversion. Note that the type is
gawk_int_t and NOT necessarily long.
|
