The libc.prx that titles ship (PPSA05916, PPSA07923) stores every NaN a
floating conversion reads as the positive default quiet NaN, whatever
its sign or payload: -nan gives 0x7fc00000 for %f, 0x7ff8000000000000
for %lf and 0x7fff:c000000000000000 for %Lf. The host scanners keep the
sign. The sscanf_s scanner now rewrites a stored NaN, ten bytes for %Lf.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The narrow sscanf of titles' libc.prx (PPSA07923 and PPSA05916 agree)
reads a %l[ set left to right: X-Y is a byte range when a - follows a
member and is not last, a reversed range matches nothing, and a - first,
last or after a range is a member. InWideScanset made a reversed range's
first end a member and chained a following -, so %l[z-a] took z and
%l[a-a-c] took b. It now follows the rule swscanf's InSet uses.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Titles' libc.prx (PPSA07923, PPSA05916, PPSA13934) reads %ls, %lc, %l[
and %S by passing each byte to mbrtowc with the state kept for the
field, so in the C locale any byte from 0x80 up ends the conversion:
- a UTF-8 lead byte is held, and the next byte the field takes fails it
with EILSEQ as a matching failure;
- a field that ends while a byte is held is an input failure;
- any other byte mbrtowc refuses is an input failure as the field's
first byte and a matching failure after it;
- the units read before a failure stay stored without a terminator.
Widths count bytes, scanset members are compared as bytes, and %C is
not a conversion there, so it ends the scan.
The guest scanner decoded UTF-8 instead and threw for anything else, and
Windows handed these conversions to the host, which stored each byte as
one unit and threw for %S and %C. The scanner now does the above on every
platform, and Windows sends formats with a wide conversion to it, as
Linux does. Wide detection now follows the scanner's own rule, so %hS and
%lS count. The wide tests run on Windows too and expect the titles'
results; the UTF-8 cases they had are replaced by measured ones.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The libc.prx that titles ship (PPSA05916, PPSA07923) does not take a
0x prefix with no hex digit after it as 0. A %x, %X, %i or %p field
fails as an input failure when the input or width ends after the x
and no sign precedes it, and as a matching failure otherwise; a
floating field needs a hex digit, or a dot and a hex digit. Floating
fields set ERANGE only when a decimal field is out of long double
range, so %e of 1e40 and %a of 0x1p200 leave errno alone.
sscanf_s and vsscanf_s now check the prefix before handing the field
to the host scanner and decide errno from the scanned text. The
stream forms cannot look three characters ahead and are unchanged.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The sscanf_s of titles' libc.prx (PPSA07923 and PPSA05916 agree) reads a
%d, %i, %o, %u, %x or %X with no length, h or hh as a 64-bit value and
keeps its low bits: %x of 61439ebdf stores 0x1439ebdf and leaves errno
alone. ScanSource handed the directive to the host at the destination's
width, which saturated to 0xffffffff and set ERANGE. It now scans those
conversions into a 64-bit temporary and stores the low bytes, as the
plain Windows sscanf path already does. Out of 64-bit range the host
still saturates with ERANGE, as the titles do.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The narrow sscanf and sscanf_s of titles' libc.prx (PPSA07923 and
PPSA05916 agree) differ from the host scanner Windows used:
- scanset ranges are read left to right: X-Y is a range when a - follows
a member and is not last, a reversed range matches nothing, a - first,
last or after a range is a member, and a ] that opens the set can
start a range; an unterminated set ends the scan as a matching failure;
- %s, %c and %[ read narrow with any length but l, where the host read
%lls as wide;
- an unknown conversion such as %y is a matching failure, even at the
end of the input, where it threw or returned EOF.
Scansets are now rewritten as explicit byte lists, so the host's range
rules never apply, the length of s, c and [ is dropped unless it is l,
and an unknown conversion ends the scan.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
%Ld formats an int once L is ignored, so the test passes a negative int
instead of an unsigned value, and %Ln writes an int, so the test counts
into an int followed by a guard instead of reading a long long.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The printf family threw for a length modifier its conversion does not
take, such as %Ld, %hhc, %Ls, %hf or %hp, and the wide family returned
-1. The libc.prx bundled with titles ignores such a modifier: L on an
integer conversion or %n (an int), any length but l on %c and %s, any
but L on a floating conversion, and any on %p. Both formatters now drop
it before converting, and on Linux such a format goes to the guest
formatter. The _s forms, unmeasured, still throw.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A long double %La went to the host, which on Windows (MinGW) prints the
64-bit significand with its integer bit as the leading digit:
0xc.ccccccccccccccdp-7 for 0.1L. The libc.prx that titles ship
normalises it like a double, with a leading 1 and the remaining 63 bits
as 16 hexadecimal digits, 0x1.999999999999999ap-4, and rounds a shorter
precision half up with the carry kept in the leading digit. Share
HexFloat's digit and rounding code between double and long double and
use it for finite long double %a and %A in the narrow and wide families.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The printf family formatted floating-point conversions through the host,
which prints every exact digit and rounds ties to even. The libc.prx
that titles ship keeps the exact digits only down to a position derived
from the binary exponent and prints zeros after them, rounds half up on
the kept digits, and rounds an exact tie down once that exponent
estimate reaches 8. So %.0f of 2.5 is 3 and %.1f of 0.25 is 0.3 there,
and %.20f of 0.1 is 0.10000000000000000555.
Format finite double and long double e, f and g conversions with those
rules from exactly generated digits, on every host. On 600,000 random
formats every output with up to 17 significant digits now matches the
copies bundled with PPSA07923, PPSA05916 and PPSA13934; longer outputs
still differ where the console's extended-precision digits end one unit
lower.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The Windows printf family handed %a and %A to the host, whose UCRT prints
a subnormal unnormalised (0x0.0000000000001p-1022) and pads with zeros
even when the - flag is given (%-010a of 1.5 gave 0x001.8p+0). The
libc.prx that titles ship normalises subnormals (0x1p-1074), rounds a
shorter precision half up without renormalising a carry, and lets -
override 0, as C requires.
Finite doubles are now formatted here, checked against three titles'
copies of libc.prx run side by side.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Keep #3987's q, D, O, U, width, '-' and trailing '%' handling for
unknown conversions and fold it into this branch's measured path:
'#' and precision are ignored as measured, '+', space, apostrophe
and '0' with a width still throw.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The unknown-conversion branch printed %m, positional arguments, the grouping
flag and modifiers in an unusual order as literal characters. Only characters
FreeBSD's vfprintf prints as themselves are now printed.
(cherry picked from commit 0106858d3d39f83eb2234a5f581d5bf5260521a0)
The Windows printf family threw on the q length modifier, the D, O and U
conversions, an unknown conversion character and a format ending in a percent
sign, where FreeBSD's vfprintf formats the first three and prints the unknown
character itself or stops at the end of the format.
(cherry picked from commit 0e3b2896d228c601cefc992b2895723188ebf0fa)
The console's own sscanf, run natively from three titles' libc.prx, returns 0 for %qd and stores nothing, where FreeBSD 11.4's source reads q as ll. A q directive now ends the scan there, and the count of conversions before it is returned. The test adds the console's rows for wrapping narrow conversions, saturating 64-bit ones with ERANGE, and %qd.
Measured through libSceLibcInternal.sprx's snprintf in a payload process on a PS5 (firmware 13.60): any size is accepted, a NaN keeps its sign and the + and space flags, each UTF-16 unit of %ls and %lc is encoded alone, positions above 99 work, and an unknown conversion such as %m prints % and its letter.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
JJenkx