ww/lex: fold float literals through strconv.stof64 — 1-ULP cs≠ww class (#62)
wwstage's parsef64 (naive i64-accumulator + pow-10 fold) diverged from cstage's strtod: >19-digit mantissas overflowed the accumulator (sign-bit garbage), DBL_MIN was +1 ULP, DBL_MAX -2 ULP — the #59.10 ratchet pin. C-strtod oracle confirms cstage correctly rounded on every vector, so wwstage aligns to it by dogfooding strconv.stof64 (correctly-rounded decimal engine, already imported by lex.ww). Overflow literals now reject in both stages (stof64 overflow -> errat, mirroring ERANGE). Fix + #59.10 M_DIVERGE->M_ID graduation + pins land together per the ratchet's designed flow (the gate trips loud demanding graduation): oracle-pinned vectors in toktest.ww floatfold_cases (lexer-unit) and 989_floatlit_run (compiler fold: runtime bits + byte-id + overflow reject parity). Retained subnormal accept-set asymmetry filed as task #21, documented at the lexnum site.
This commit is contained in:
5
Makefile
5
Makefile
@@ -427,6 +427,7 @@ TESTS = $(BIN)/test_smoke $(BIN)/test_lex $(BIN)/test_parse $(BIN)/test_check \
|
|||||||
$(BIN)/test_siphash_run $(BIN)/test_sha256_run \
|
$(BIN)/test_siphash_run $(BIN)/test_sha256_run \
|
||||||
$(BIN)/test_regex_run \
|
$(BIN)/test_regex_run \
|
||||||
$(BIN)/test_lib_byteid \
|
$(BIN)/test_lib_byteid \
|
||||||
|
$(BIN)/test_floatlit_run \
|
||||||
$(BIN)/test_checked_run \
|
$(BIN)/test_checked_run \
|
||||||
$(BIN)/test_floatarr_run \
|
$(BIN)/test_floatarr_run \
|
||||||
$(BIN)/test_deref_narrow_run \
|
$(BIN)/test_deref_narrow_run \
|
||||||
@@ -1830,6 +1831,10 @@ $(BIN)/test_lib_byteid: test/wcc/989_lib_byteid.c $(BIN)/ww $(BIN)/w6c \
|
|||||||
$(BIN)/w6c_ww $(BIN)/w6a $(BIN)/w6l $(LIB)/libwwrt.a | $(BIN)
|
$(BIN)/w6c_ww $(BIN)/w6a $(BIN)/w6l $(LIB)/libwwrt.a | $(BIN)
|
||||||
$(CC) $(CFLAGS) -o $@ $<
|
$(CC) $(CFLAGS) -o $@ $<
|
||||||
|
|
||||||
|
$(BIN)/test_floatlit_run: test/wcc/989_floatlit_run.c $(BIN)/ww $(BIN)/w6c \
|
||||||
|
$(BIN)/w6c_ww $(BIN)/w6a $(BIN)/w6l $(LIB)/libwwrt.a | $(BIN)
|
||||||
|
$(CC) $(CFLAGS) -o $@ $<
|
||||||
|
|
||||||
$(BIN)/test_bufio_run: test/wcc/998_bufio_run.c $(BIN)/ww $(BIN)/w6c \
|
$(BIN)/test_bufio_run: test/wcc/998_bufio_run.c $(BIN)/ww $(BIN)/w6c \
|
||||||
$(BIN)/w6a $(BIN)/w6l $(LIB)/libwwrt.a | $(BIN)
|
$(BIN)/w6a $(BIN)/w6l $(LIB)/libwwrt.a | $(BIN)
|
||||||
$(CC) $(CFLAGS) -o $@ $<
|
$(CC) $(CFLAGS) -o $@ $<
|
||||||
|
|||||||
@@ -298,82 +298,6 @@ fn scanexp(l: *lex) void = {
|
|||||||
};
|
};
|
||||||
};
|
};
|
||||||
|
|
||||||
// parsef64 — minimal decimal-float parser. Reads digits[.digits][eE[+-]digits]
|
|
||||||
// from the first `n` bytes of `s` (no leading sign — the lexer emits
|
|
||||||
// the unary minus as a separate token). The result rounds to the
|
|
||||||
// nearest f64 only via the trailing pow-10 multiply; this matches
|
|
||||||
// `strtod` to 1 ULP on typical literals and is good enough for the
|
|
||||||
// wwstage's own use (no float literals appear in the bootstrap
|
|
||||||
// source). Anything past `n` or non-digit is silently ignored.
|
|
||||||
fn parsef64(s: *u8, n: u64) f64 = {
|
|
||||||
let i: u64 = 0u64;
|
|
||||||
let intp: i64 = 0i64;
|
|
||||||
for (i < n) {
|
|
||||||
let b: u8 = s[i];
|
|
||||||
if (b < 48u8) { break; };
|
|
||||||
if (b > 57u8) { break; };
|
|
||||||
intp = intp * 10i64 + (b - 48u8): i64;
|
|
||||||
i += 1u64;
|
|
||||||
};
|
|
||||||
let frac: i64 = 0i64;
|
|
||||||
let fscale: i64 = 1i64;
|
|
||||||
if (i < n) {
|
|
||||||
if (s[i] == '.') {
|
|
||||||
i += 1u64;
|
|
||||||
for (i < n) {
|
|
||||||
let b: u8 = s[i];
|
|
||||||
if (b < 48u8) { break; };
|
|
||||||
if (b > 57u8) { break; };
|
|
||||||
frac = frac * 10i64 + (b - 48u8): i64;
|
|
||||||
fscale = fscale * 10i64;
|
|
||||||
i += 1u64;
|
|
||||||
};
|
|
||||||
};
|
|
||||||
};
|
|
||||||
let exp: i32 = 0;
|
|
||||||
let expneg: bool = false;
|
|
||||||
if (i < n) {
|
|
||||||
let e: u8 = s[i];
|
|
||||||
if (e == 'e' || e == 'E') {
|
|
||||||
i += 1u64;
|
|
||||||
if (i < n) {
|
|
||||||
if (s[i] == '-') {
|
|
||||||
expneg = true;
|
|
||||||
i += 1u64;
|
|
||||||
} else { if (s[i] == '+') {
|
|
||||||
i += 1u64;
|
|
||||||
};};
|
|
||||||
};
|
|
||||||
for (i < n) {
|
|
||||||
let b: u8 = s[i];
|
|
||||||
if (b < 48u8) { break; };
|
|
||||||
if (b > 57u8) { break; };
|
|
||||||
exp = exp * 10 + (b - 48u8): i32;
|
|
||||||
i += 1u64;
|
|
||||||
};
|
|
||||||
};
|
|
||||||
};
|
|
||||||
let result: f64 = intp: f64;
|
|
||||||
if (frac != 0i64) {
|
|
||||||
result = result + (frac: f64) / (fscale: f64);
|
|
||||||
};
|
|
||||||
if (exp != 0) {
|
|
||||||
// Use int-to-float casts so this file stays free of float
|
|
||||||
// literals — 990's wwdump diff relies on lib/ww/lex/lex.ww
|
|
||||||
// tokenising identically through C and ww, and the C dumper
|
|
||||||
// %g-formats TK_FLOAT.fval while the ww dumper currently
|
|
||||||
// skips it. Hiding the constants behind casts keeps both
|
|
||||||
// sides emitting `FLOAT` with no payload.
|
|
||||||
let factor: f64 = 1: f64;
|
|
||||||
let ten: f64 = 10: f64;
|
|
||||||
let k: i32 = 0;
|
|
||||||
for (k < exp) { factor = factor * ten; k += 1; };
|
|
||||||
if (expneg) { result = result / factor; }
|
|
||||||
else { result = result * factor; };
|
|
||||||
};
|
|
||||||
return result;
|
|
||||||
};
|
|
||||||
|
|
||||||
fn lexnum(l: *lex, start: *pos, out: *tok) void = {
|
fn lexnum(l: *lex, start: *pos, out: *tok) void = {
|
||||||
out.kind = tkind.TK_INT;
|
out.kind = tkind.TK_INT;
|
||||||
out.file = start.file;
|
out.file = start.file;
|
||||||
@@ -451,7 +375,30 @@ fn lexnum(l: *lex, start: *pos, out: *tok) void = {
|
|||||||
i += 1u64;
|
i += 1u64;
|
||||||
};
|
};
|
||||||
clean[j] = 0u8;
|
clean[j] = 0u8;
|
||||||
let fv: f64 = parsef64(clean.ptr, j);
|
let cleanv: str;
|
||||||
|
cleanv.ptr = clean.ptr;
|
||||||
|
cleanv.len = j: i32;
|
||||||
|
// strconv's correctly-rounded decimal engine — cstage folds
|
||||||
|
// via strtod, and a leaner pow-10 fold here was 1-2 ULP off
|
||||||
|
// on long-mantissa/extreme literals (cs≠ww DATA bits, #62).
|
||||||
|
// `0: f64` cast, not a 0.0 literal: 990's wwdump diff relies
|
||||||
|
// on this file tokenising identically through C and ww, and
|
||||||
|
// the C dumper %g-formats TK_FLOAT.fval while the ww dumper
|
||||||
|
// skips it.
|
||||||
|
// Retained divergence (task #21): SUBNORMAL literals are
|
||||||
|
// accepted here correctly-rounded (Hare stof semantics)
|
||||||
|
// but rejected by cstage (glibc strtod flags partial
|
||||||
|
// underflow with ERANGE).
|
||||||
|
let fv: f64 = 0: f64;
|
||||||
|
match (strconv.stof64(cleanv, strconv.base.DEC)) {
|
||||||
|
case let v: f64 => { fv = v; };
|
||||||
|
case let e: strconv.invalid => {
|
||||||
|
errat(l, start, "bad float literal");
|
||||||
|
};
|
||||||
|
case let e: strconv.overflow => {
|
||||||
|
errat(l, start, "bad float literal");
|
||||||
|
};
|
||||||
|
};
|
||||||
out.fval = fv;
|
out.fval = fv;
|
||||||
// Stash the IEEE bits in uval — cgen consumers read floats
|
// Stash the IEEE bits in uval — cgen consumers read floats
|
||||||
// as integers (n.uval) to avoid an SSE round-trip when
|
// as integers (n.uval) to avoid an SSE round-trip when
|
||||||
|
|||||||
@@ -272,9 +272,62 @@ fn checkprint(fd: i32, t: *tok, want: str) void = {
|
|||||||
if (os.remove("/tmp/ww_s9_tok.tmp") != 0) { fail(); };
|
if (os.remove("/tmp/ww_s9_tok.tmp") != 0) { fail(); };
|
||||||
};
|
};
|
||||||
|
|
||||||
|
fn checkfloat(src: str, want: u64) void = {
|
||||||
|
let l: lex;
|
||||||
|
lexinit(&l, "t", src.ptr, src.len: u64);
|
||||||
|
let t: tok;
|
||||||
|
lexnext(&l, &t);
|
||||||
|
if (t.kind != tkind.TK_FLOAT) { fail(); };
|
||||||
|
if (t.uval != want) { fail(); };
|
||||||
|
};
|
||||||
|
|
||||||
|
// #62 pin: lexnum's float fold routes through strconv.stof64 and must
|
||||||
|
// produce the IEEE-754 correctly-rounded bits cstage gets from strtod
|
||||||
|
// — any rounding slip is a cs≠ww DATA divergence. Vectors pinned
|
||||||
|
// against a C strtod oracle. Rows cover the classes the retired
|
||||||
|
// pow-10 fold got wrong: >19-digit mantissas (its i64 accumulator
|
||||||
|
// overflowed), DBL_MIN/DBL_MAX extremes, and the decimal-fraction
|
||||||
|
// 1-ULP double-rounding cases; plus halfway-to-even, exponent forms,
|
||||||
|
// the underscore strip, the 53-digit exact-halfway pair at the 2^-53
|
||||||
|
// boundary (tie rounds to even, tie+1 rounds up — also the only
|
||||||
|
// >19-digit FRACTION rows), and an exact power of two.
|
||||||
|
@test fn floatfold_cases() void = {
|
||||||
|
signalled = 400; checkfloat("1.0000000000000002", 0x3FF0000000000001u64);
|
||||||
|
signalled = 401; checkfloat("9007199254740993.0", 0x4340000000000000u64);
|
||||||
|
signalled = 402; checkfloat("1.2345e67", 0x4DDD4E421712C0B7u64);
|
||||||
|
signalled = 403; checkfloat("0.1", 0x3FB999999999999Au64);
|
||||||
|
signalled = 404; checkfloat("1.1", 0x3FF199999999999Au64);
|
||||||
|
signalled = 405;
|
||||||
|
checkfloat("123456789012345678901234567890.0", 0x45F8EE90FF6C373Eu64);
|
||||||
|
signalled = 406;
|
||||||
|
checkfloat("2.2250738585072014e-308", 0x0010000000000000u64);
|
||||||
|
signalled = 407; checkfloat("0.3", 0x3FD3333333333333u64);
|
||||||
|
signalled = 408; checkfloat("3.141592653589793", 0x400921FB54442D18u64);
|
||||||
|
signalled = 409;
|
||||||
|
checkfloat("1.7976931348623157e308", 0x7FEFFFFFFFFFFFFFu64);
|
||||||
|
signalled = 410;
|
||||||
|
checkfloat("1.7976931348623158e308", 0x7FEFFFFFFFFFFFFFu64);
|
||||||
|
signalled = 411; checkfloat("7.2057594037927933e16", 0x4370000000000000u64);
|
||||||
|
signalled = 412;
|
||||||
|
checkfloat("1000000000000000000000.0", 0x444B1AE4D6E2EF50u64);
|
||||||
|
signalled = 413; checkfloat("1_000.5", 0x408F440000000000u64);
|
||||||
|
signalled = 414;
|
||||||
|
checkfloat("1.00000000000000011102230246251565404236316680908203125",
|
||||||
|
0x3FF0000000000000u64);
|
||||||
|
signalled = 415;
|
||||||
|
checkfloat("1.00000000000000011102230246251565404236316680908203126",
|
||||||
|
0x3FF0000000000001u64);
|
||||||
|
signalled = 416; checkfloat("4503599627370497.5", 0x4330000000000002u64);
|
||||||
|
signalled = 417; checkfloat("0.5", 0x3FE0000000000000u64);
|
||||||
|
signalled = 418; checkfloat("1.0e308", 0x7FE1CCF385EBC8A0u64);
|
||||||
|
signalled = 419;
|
||||||
|
checkfloat("2.225073858507202e-308", 0x0010000000000001u64);
|
||||||
|
};
|
||||||
|
|
||||||
export fn main() i32 = {
|
export fn main() i32 = {
|
||||||
signalled = 1; tokname_cases();
|
signalled = 1; tokname_cases();
|
||||||
signalled = 2; kwlookup_cases();
|
signalled = 2; kwlookup_cases();
|
||||||
signalled = 3; tokprint_cases();
|
signalled = 3; tokprint_cases();
|
||||||
|
signalled = 4; floatfold_cases();
|
||||||
return 0;
|
return 0;
|
||||||
};
|
};
|
||||||
|
|||||||
@@ -7017,82 +7017,6 @@ fn scanexp(l: *lex) void = {
|
|||||||
};
|
};
|
||||||
};
|
};
|
||||||
|
|
||||||
// parsef64 — minimal decimal-float parser. Reads digits[.digits][eE[+-]digits]
|
|
||||||
// from the first `n` bytes of `s` (no leading sign — the lexer emits
|
|
||||||
// the unary minus as a separate token). The result rounds to the
|
|
||||||
// nearest f64 only via the trailing pow-10 multiply; this matches
|
|
||||||
// `strtod` to 1 ULP on typical literals and is good enough for the
|
|
||||||
// wwstage's own use (no float literals appear in the bootstrap
|
|
||||||
// source). Anything past `n` or non-digit is silently ignored.
|
|
||||||
fn parsef64(s: *u8, n: u64) f64 = {
|
|
||||||
let i: u64 = 0u64;
|
|
||||||
let intp: i64 = 0i64;
|
|
||||||
for (i < n) {
|
|
||||||
let b: u8 = s[i];
|
|
||||||
if (b < 48u8) { break; };
|
|
||||||
if (b > 57u8) { break; };
|
|
||||||
intp = intp * 10i64 + (b - 48u8): i64;
|
|
||||||
i += 1u64;
|
|
||||||
};
|
|
||||||
let frac: i64 = 0i64;
|
|
||||||
let fscale: i64 = 1i64;
|
|
||||||
if (i < n) {
|
|
||||||
if (s[i] == '.') {
|
|
||||||
i += 1u64;
|
|
||||||
for (i < n) {
|
|
||||||
let b: u8 = s[i];
|
|
||||||
if (b < 48u8) { break; };
|
|
||||||
if (b > 57u8) { break; };
|
|
||||||
frac = frac * 10i64 + (b - 48u8): i64;
|
|
||||||
fscale = fscale * 10i64;
|
|
||||||
i += 1u64;
|
|
||||||
};
|
|
||||||
};
|
|
||||||
};
|
|
||||||
let exp: i32 = 0;
|
|
||||||
let expneg: bool = false;
|
|
||||||
if (i < n) {
|
|
||||||
let e: u8 = s[i];
|
|
||||||
if (e == 'e' || e == 'E') {
|
|
||||||
i += 1u64;
|
|
||||||
if (i < n) {
|
|
||||||
if (s[i] == '-') {
|
|
||||||
expneg = true;
|
|
||||||
i += 1u64;
|
|
||||||
} else { if (s[i] == '+') {
|
|
||||||
i += 1u64;
|
|
||||||
};};
|
|
||||||
};
|
|
||||||
for (i < n) {
|
|
||||||
let b: u8 = s[i];
|
|
||||||
if (b < 48u8) { break; };
|
|
||||||
if (b > 57u8) { break; };
|
|
||||||
exp = exp * 10 + (b - 48u8): i32;
|
|
||||||
i += 1u64;
|
|
||||||
};
|
|
||||||
};
|
|
||||||
};
|
|
||||||
let result: f64 = intp: f64;
|
|
||||||
if (frac != 0i64) {
|
|
||||||
result = result + (frac: f64) / (fscale: f64);
|
|
||||||
};
|
|
||||||
if (exp != 0) {
|
|
||||||
// Use int-to-float casts so this file stays free of float
|
|
||||||
// literals — 990's wwdump diff relies on lib/ww/lex/lex.ww
|
|
||||||
// tokenising identically through C and ww, and the C dumper
|
|
||||||
// %g-formats TK_FLOAT.fval while the ww dumper currently
|
|
||||||
// skips it. Hiding the constants behind casts keeps both
|
|
||||||
// sides emitting `FLOAT` with no payload.
|
|
||||||
let factor: f64 = 1: f64;
|
|
||||||
let ten: f64 = 10: f64;
|
|
||||||
let k: i32 = 0;
|
|
||||||
for (k < exp) { factor = factor * ten; k += 1; };
|
|
||||||
if (expneg) { result = result / factor; }
|
|
||||||
else { result = result * factor; };
|
|
||||||
};
|
|
||||||
return result;
|
|
||||||
};
|
|
||||||
|
|
||||||
fn lexnum(l: *lex, start: *pos, out: *tok) void = {
|
fn lexnum(l: *lex, start: *pos, out: *tok) void = {
|
||||||
out.kind = tkind.TK_INT;
|
out.kind = tkind.TK_INT;
|
||||||
out.file = start.file;
|
out.file = start.file;
|
||||||
@@ -7170,7 +7094,30 @@ fn lexnum(l: *lex, start: *pos, out: *tok) void = {
|
|||||||
i += 1u64;
|
i += 1u64;
|
||||||
};
|
};
|
||||||
clean[j] = 0u8;
|
clean[j] = 0u8;
|
||||||
let fv: f64 = parsef64(clean.ptr, j);
|
let cleanv: str;
|
||||||
|
cleanv.ptr = clean.ptr;
|
||||||
|
cleanv.len = j: i32;
|
||||||
|
// strconv's correctly-rounded decimal engine — cstage folds
|
||||||
|
// via strtod, and a leaner pow-10 fold here was 1-2 ULP off
|
||||||
|
// on long-mantissa/extreme literals (cs≠ww DATA bits, #62).
|
||||||
|
// `0: f64` cast, not a 0.0 literal: 990's wwdump diff relies
|
||||||
|
// on this file tokenising identically through C and ww, and
|
||||||
|
// the C dumper %g-formats TK_FLOAT.fval while the ww dumper
|
||||||
|
// skips it.
|
||||||
|
// Retained divergence (task #21): SUBNORMAL literals are
|
||||||
|
// accepted here correctly-rounded (Hare stof semantics)
|
||||||
|
// but rejected by cstage (glibc strtod flags partial
|
||||||
|
// underflow with ERANGE).
|
||||||
|
let fv: f64 = 0: f64;
|
||||||
|
match (strconv.stof64(cleanv, strconv.base.DEC)) {
|
||||||
|
case let v: f64 => { fv = v; };
|
||||||
|
case let e: strconv.invalid => {
|
||||||
|
errat(l, start, "bad float literal");
|
||||||
|
};
|
||||||
|
case let e: strconv.overflow => {
|
||||||
|
errat(l, start, "bad float literal");
|
||||||
|
};
|
||||||
|
};
|
||||||
out.fval = fv;
|
out.fval = fv;
|
||||||
// Stash the IEEE bits in uval — cgen consumers read floats
|
// Stash the IEEE bits in uval — cgen consumers read floats
|
||||||
// as integers (n.uval) to avoid an SSE round-trip when
|
// as integers (n.uval) to avoid an SSE round-trip when
|
||||||
|
|||||||
@@ -7017,82 +7017,6 @@ fn scanexp(l: *lex) void = {
|
|||||||
};
|
};
|
||||||
};
|
};
|
||||||
|
|
||||||
// parsef64 — minimal decimal-float parser. Reads digits[.digits][eE[+-]digits]
|
|
||||||
// from the first `n` bytes of `s` (no leading sign — the lexer emits
|
|
||||||
// the unary minus as a separate token). The result rounds to the
|
|
||||||
// nearest f64 only via the trailing pow-10 multiply; this matches
|
|
||||||
// `strtod` to 1 ULP on typical literals and is good enough for the
|
|
||||||
// wwstage's own use (no float literals appear in the bootstrap
|
|
||||||
// source). Anything past `n` or non-digit is silently ignored.
|
|
||||||
fn parsef64(s: *u8, n: u64) f64 = {
|
|
||||||
let i: u64 = 0u64;
|
|
||||||
let intp: i64 = 0i64;
|
|
||||||
for (i < n) {
|
|
||||||
let b: u8 = s[i];
|
|
||||||
if (b < 48u8) { break; };
|
|
||||||
if (b > 57u8) { break; };
|
|
||||||
intp = intp * 10i64 + (b - 48u8): i64;
|
|
||||||
i += 1u64;
|
|
||||||
};
|
|
||||||
let frac: i64 = 0i64;
|
|
||||||
let fscale: i64 = 1i64;
|
|
||||||
if (i < n) {
|
|
||||||
if (s[i] == '.') {
|
|
||||||
i += 1u64;
|
|
||||||
for (i < n) {
|
|
||||||
let b: u8 = s[i];
|
|
||||||
if (b < 48u8) { break; };
|
|
||||||
if (b > 57u8) { break; };
|
|
||||||
frac = frac * 10i64 + (b - 48u8): i64;
|
|
||||||
fscale = fscale * 10i64;
|
|
||||||
i += 1u64;
|
|
||||||
};
|
|
||||||
};
|
|
||||||
};
|
|
||||||
let exp: i32 = 0;
|
|
||||||
let expneg: bool = false;
|
|
||||||
if (i < n) {
|
|
||||||
let e: u8 = s[i];
|
|
||||||
if (e == 'e' || e == 'E') {
|
|
||||||
i += 1u64;
|
|
||||||
if (i < n) {
|
|
||||||
if (s[i] == '-') {
|
|
||||||
expneg = true;
|
|
||||||
i += 1u64;
|
|
||||||
} else { if (s[i] == '+') {
|
|
||||||
i += 1u64;
|
|
||||||
};};
|
|
||||||
};
|
|
||||||
for (i < n) {
|
|
||||||
let b: u8 = s[i];
|
|
||||||
if (b < 48u8) { break; };
|
|
||||||
if (b > 57u8) { break; };
|
|
||||||
exp = exp * 10 + (b - 48u8): i32;
|
|
||||||
i += 1u64;
|
|
||||||
};
|
|
||||||
};
|
|
||||||
};
|
|
||||||
let result: f64 = intp: f64;
|
|
||||||
if (frac != 0i64) {
|
|
||||||
result = result + (frac: f64) / (fscale: f64);
|
|
||||||
};
|
|
||||||
if (exp != 0) {
|
|
||||||
// Use int-to-float casts so this file stays free of float
|
|
||||||
// literals — 990's wwdump diff relies on lib/ww/lex/lex.ww
|
|
||||||
// tokenising identically through C and ww, and the C dumper
|
|
||||||
// %g-formats TK_FLOAT.fval while the ww dumper currently
|
|
||||||
// skips it. Hiding the constants behind casts keeps both
|
|
||||||
// sides emitting `FLOAT` with no payload.
|
|
||||||
let factor: f64 = 1: f64;
|
|
||||||
let ten: f64 = 10: f64;
|
|
||||||
let k: i32 = 0;
|
|
||||||
for (k < exp) { factor = factor * ten; k += 1; };
|
|
||||||
if (expneg) { result = result / factor; }
|
|
||||||
else { result = result * factor; };
|
|
||||||
};
|
|
||||||
return result;
|
|
||||||
};
|
|
||||||
|
|
||||||
fn lexnum(l: *lex, start: *pos, out: *tok) void = {
|
fn lexnum(l: *lex, start: *pos, out: *tok) void = {
|
||||||
out.kind = tkind.TK_INT;
|
out.kind = tkind.TK_INT;
|
||||||
out.file = start.file;
|
out.file = start.file;
|
||||||
@@ -7170,7 +7094,30 @@ fn lexnum(l: *lex, start: *pos, out: *tok) void = {
|
|||||||
i += 1u64;
|
i += 1u64;
|
||||||
};
|
};
|
||||||
clean[j] = 0u8;
|
clean[j] = 0u8;
|
||||||
let fv: f64 = parsef64(clean.ptr, j);
|
let cleanv: str;
|
||||||
|
cleanv.ptr = clean.ptr;
|
||||||
|
cleanv.len = j: i32;
|
||||||
|
// strconv's correctly-rounded decimal engine — cstage folds
|
||||||
|
// via strtod, and a leaner pow-10 fold here was 1-2 ULP off
|
||||||
|
// on long-mantissa/extreme literals (cs≠ww DATA bits, #62).
|
||||||
|
// `0: f64` cast, not a 0.0 literal: 990's wwdump diff relies
|
||||||
|
// on this file tokenising identically through C and ww, and
|
||||||
|
// the C dumper %g-formats TK_FLOAT.fval while the ww dumper
|
||||||
|
// skips it.
|
||||||
|
// Retained divergence (task #21): SUBNORMAL literals are
|
||||||
|
// accepted here correctly-rounded (Hare stof semantics)
|
||||||
|
// but rejected by cstage (glibc strtod flags partial
|
||||||
|
// underflow with ERANGE).
|
||||||
|
let fv: f64 = 0: f64;
|
||||||
|
match (strconv.stof64(cleanv, strconv.base.DEC)) {
|
||||||
|
case let v: f64 => { fv = v; };
|
||||||
|
case let e: strconv.invalid => {
|
||||||
|
errat(l, start, "bad float literal");
|
||||||
|
};
|
||||||
|
case let e: strconv.overflow => {
|
||||||
|
errat(l, start, "bad float literal");
|
||||||
|
};
|
||||||
|
};
|
||||||
out.fval = fv;
|
out.fval = fv;
|
||||||
// Stash the IEEE bits in uval — cgen consumers read floats
|
// Stash the IEEE bits in uval — cgen consumers read floats
|
||||||
// as integers (n.uval) to avoid an SSE round-trip when
|
// as integers (n.uval) to avoid an SSE round-trip when
|
||||||
|
|||||||
205
test/wcc/989_floatlit_run.c
Normal file
205
test/wcc/989_floatlit_run.c
Normal file
@@ -0,0 +1,205 @@
|
|||||||
|
/*
|
||||||
|
* 989_floatlit_run — runtime + byte-id net for #62: the two stages
|
||||||
|
* folded float LITERALS differently. cstage folds via strtod
|
||||||
|
* (correctly rounded); the wwstage lexer used a pow-10 accumulation
|
||||||
|
* fold that was 1-2 ULP off on decimal fractions, overflowed its i64
|
||||||
|
* accumulator past 19 mantissa digits, and missed DBL_MIN/DBL_MAX by
|
||||||
|
* up to 2 ULP — a pinned cs≠ww DATA divergence (989 ratchet #59.10).
|
||||||
|
* Fix: lexnum folds through strconv.stof64 (the Hare-ported
|
||||||
|
* correctly-rounded decimal engine).
|
||||||
|
*
|
||||||
|
* Fixture 1 (vectors): each row reads back a literal's IEEE bits via
|
||||||
|
* a *u64 reinterpret and compares against the C-strtod-oracle bit
|
||||||
|
* pattern; the exit code pinpoints the failing row. Carries (a)
|
||||||
|
* cstage `ww build` + run asserting exit 0 (rule-10: convergence
|
||||||
|
* targets the runtime-CORRECT side) and (b) w6c vs w6c_ww `.s` cmp
|
||||||
|
* (byte-id — the wwstage fold itself).
|
||||||
|
*
|
||||||
|
* Fixture 2 (overflow): 1.7976931348623159e308 rounds above DBL_MAX —
|
||||||
|
* strtod sets ERANGE so cstage rejects; wwstage must reject too
|
||||||
|
* (stof64 overflow → errat). Both compilers must exit non-zero.
|
||||||
|
*
|
||||||
|
* Retained divergence (task #21): a SUBNORMAL literal (e.g.
|
||||||
|
* 2.2250738585072011e-308) is rejected by cstage (glibc strtod flags
|
||||||
|
* partial underflow with ERANGE) but accepted correctly-rounded by
|
||||||
|
* wwstage (Hare stof semantics) — accept-set asymmetry only, not a
|
||||||
|
* bits divergence; pre-existing on master (old parsef64 accepted it
|
||||||
|
* with garbage bits).
|
||||||
|
*
|
||||||
|
* 9xx is full; shares the 989 prefix per the 989_sha256 precedent.
|
||||||
|
*/
|
||||||
|
#include <stdio.h>
|
||||||
|
#include <stdlib.h>
|
||||||
|
#include <unistd.h>
|
||||||
|
#include <sys/stat.h>
|
||||||
|
#include <sys/wait.h>
|
||||||
|
|
||||||
|
static const char *vectors_src =
|
||||||
|
"package main;\n"
|
||||||
|
"fn bits(v: f64) u64 = {\n"
|
||||||
|
" let x: f64 = v;\n"
|
||||||
|
" let p: *u64 = (&x): *u64;\n"
|
||||||
|
" return *p;\n"
|
||||||
|
"};\n"
|
||||||
|
"export fn main() i32 = {\n"
|
||||||
|
" if (bits(1.0000000000000002) != 0x3FF0000000000001u64) { return 1; };\n"
|
||||||
|
" if (bits(9007199254740993.0) != 0x4340000000000000u64) { return 2; };\n"
|
||||||
|
" if (bits(1.2345e67) != 0x4DDD4E421712C0B7u64) { return 3; };\n"
|
||||||
|
" if (bits(0.1) != 0x3FB999999999999Au64) { return 4; };\n"
|
||||||
|
" if (bits(1.1) != 0x3FF199999999999Au64) { return 5; };\n"
|
||||||
|
" if (bits(123456789012345678901234567890.0) != 0x45F8EE90FF6C373Eu64) { return 6; };\n"
|
||||||
|
" if (bits(2.2250738585072014e-308) != 0x0010000000000000u64) { return 7; };\n"
|
||||||
|
" if (bits(0.3) != 0x3FD3333333333333u64) { return 8; };\n"
|
||||||
|
" if (bits(3.141592653589793) != 0x400921FB54442D18u64) { return 9; };\n"
|
||||||
|
" if (bits(1.7976931348623157e308) != 0x7FEFFFFFFFFFFFFFu64) { return 10; };\n"
|
||||||
|
" if (bits(1.7976931348623158e308) != 0x7FEFFFFFFFFFFFFFu64) { return 11; };\n"
|
||||||
|
" if (bits(7.2057594037927933e16) != 0x4370000000000000u64) { return 12; };\n"
|
||||||
|
" if (bits(1000000000000000000000.0) != 0x444B1AE4D6E2EF50u64) { return 13; };\n"
|
||||||
|
" if (bits(1_000.5) != 0x408F440000000000u64) { return 14; };\n"
|
||||||
|
" if (bits(1.00000000000000011102230246251565404236316680908203125) != 0x3FF0000000000000u64) { return 15; };\n"
|
||||||
|
" if (bits(1.00000000000000011102230246251565404236316680908203126) != 0x3FF0000000000001u64) { return 16; };\n"
|
||||||
|
" if (bits(4503599627370497.5) != 0x4330000000000002u64) { return 17; };\n"
|
||||||
|
" if (bits(0.5) != 0x3FE0000000000000u64) { return 18; };\n"
|
||||||
|
" if (bits(1.0e308) != 0x7FE1CCF385EBC8A0u64) { return 19; };\n"
|
||||||
|
" if (bits(2.225073858507202e-308) != 0x0010000000000001u64) { return 20; };\n"
|
||||||
|
" return 0;\n"
|
||||||
|
"};\n";
|
||||||
|
|
||||||
|
static const char *overflow_src =
|
||||||
|
"package main;\n"
|
||||||
|
"export fn main() i32 = {\n"
|
||||||
|
" let a: f64 = 1.7976931348623159e308;\n"
|
||||||
|
" return 0;\n"
|
||||||
|
"};\n";
|
||||||
|
|
||||||
|
static int
|
||||||
|
runwait(const char *cmd)
|
||||||
|
{
|
||||||
|
int rc = system(cmd);
|
||||||
|
if (rc == -1) return -1;
|
||||||
|
if (WIFEXITED(rc)) return WEXITSTATUS(rc);
|
||||||
|
return -1;
|
||||||
|
}
|
||||||
|
|
||||||
|
static int
|
||||||
|
slurp_eq(const char *a, const char *b)
|
||||||
|
{
|
||||||
|
FILE *fa = fopen(a, "rb");
|
||||||
|
FILE *fb = fopen(b, "rb");
|
||||||
|
if (!fa || !fb) { if (fa) fclose(fa); if (fb) fclose(fb); return -1; }
|
||||||
|
int rc = 0;
|
||||||
|
for (;;) {
|
||||||
|
int ca = fgetc(fa);
|
||||||
|
int cb = fgetc(fb);
|
||||||
|
if (ca != cb) { rc = -1; break; }
|
||||||
|
if (ca == EOF) break;
|
||||||
|
}
|
||||||
|
fclose(fa); fclose(fb);
|
||||||
|
return rc;
|
||||||
|
}
|
||||||
|
|
||||||
|
static int
|
||||||
|
writesrc(const char *path, const char *src)
|
||||||
|
{
|
||||||
|
FILE *f = fopen(path, "wb");
|
||||||
|
if (f == NULL) return -1;
|
||||||
|
fputs(src, f);
|
||||||
|
fclose(f);
|
||||||
|
return 0;
|
||||||
|
}
|
||||||
|
|
||||||
|
int
|
||||||
|
main(void)
|
||||||
|
{
|
||||||
|
const char *bin = getenv("BIN");
|
||||||
|
if (!bin) bin = "out/bin";
|
||||||
|
char absbin[1024];
|
||||||
|
if (bin[0] != '/') {
|
||||||
|
char cwd[1024];
|
||||||
|
if (getcwd(cwd, sizeof cwd) == NULL) return 1;
|
||||||
|
snprintf(absbin, sizeof absbin, "%s/%s", cwd, bin);
|
||||||
|
bin = absbin;
|
||||||
|
}
|
||||||
|
|
||||||
|
char w6c[1100], w6c_ww[1100];
|
||||||
|
snprintf(w6c, sizeof w6c, "%s/w6c", bin);
|
||||||
|
snprintf(w6c_ww, sizeof w6c_ww, "%s/w6c_ww", bin);
|
||||||
|
if (access(w6c_ww, X_OK) != 0) {
|
||||||
|
fprintf(stderr, "floatlit: w6c_ww missing — cannot run "
|
||||||
|
"the cs==ww byte-id gate (the whole point of this test)\n");
|
||||||
|
return 1;
|
||||||
|
}
|
||||||
|
|
||||||
|
int fail = 0;
|
||||||
|
char src[64], cmd[2048];
|
||||||
|
|
||||||
|
/* fixture 1: vectors — cstage build+run, then byte-id */
|
||||||
|
snprintf(src, sizeof src, "/tmp/wwflit_%d.ww", getpid());
|
||||||
|
if (writesrc(src, vectors_src) != 0) return 1;
|
||||||
|
|
||||||
|
char tmpdir[64];
|
||||||
|
snprintf(tmpdir, sizeof tmpdir, "/tmp/wwflit_%d_d", getpid());
|
||||||
|
mkdir(tmpdir, 0755);
|
||||||
|
|
||||||
|
snprintf(cmd, sizeof cmd, "cd %s && %s/ww build %s", tmpdir, bin, src);
|
||||||
|
if (runwait(cmd) != 0) {
|
||||||
|
fprintf(stderr, "floatlit: cstage build failed\n");
|
||||||
|
fail++;
|
||||||
|
} else {
|
||||||
|
char outbin[128];
|
||||||
|
snprintf(outbin, sizeof outbin, "%s/wwflit_%d", tmpdir,
|
||||||
|
getpid());
|
||||||
|
int got = runwait(outbin);
|
||||||
|
if (got != 0) {
|
||||||
|
fprintf(stderr, "floatlit: vector row %d has wrong "
|
||||||
|
"bits at runtime (cstage)\n", got);
|
||||||
|
fail++;
|
||||||
|
}
|
||||||
|
unlink(outbin);
|
||||||
|
}
|
||||||
|
rmdir(tmpdir);
|
||||||
|
|
||||||
|
char cs_s[64], ws_s[64];
|
||||||
|
snprintf(cs_s, sizeof cs_s, "/tmp/wwflit_%d_cs.s", getpid());
|
||||||
|
snprintf(ws_s, sizeof ws_s, "/tmp/wwflit_%d_ww.s", getpid());
|
||||||
|
snprintf(cmd, sizeof cmd, "%s -o %s %s 2>/dev/null", w6c, cs_s, src);
|
||||||
|
if (runwait(cmd) != 0) {
|
||||||
|
fprintf(stderr, "floatlit: w6c failed on vectors\n");
|
||||||
|
fail++;
|
||||||
|
} else {
|
||||||
|
snprintf(cmd, sizeof cmd, "%s -o %s %s 2>/dev/null",
|
||||||
|
w6c_ww, ws_s, src);
|
||||||
|
if (runwait(cmd) != 0) {
|
||||||
|
fprintf(stderr, "floatlit: w6c_ww failed on vectors\n");
|
||||||
|
fail++;
|
||||||
|
} else if (slurp_eq(cs_s, ws_s) != 0) {
|
||||||
|
fprintf(stderr, "floatlit: cstage/wwstage .s DIFFER "
|
||||||
|
"(rule-10 byte-id violation)\n");
|
||||||
|
fail++;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
unlink(src); unlink(cs_s); unlink(ws_s);
|
||||||
|
|
||||||
|
/* fixture 2: overflow literal — BOTH stages must reject */
|
||||||
|
snprintf(src, sizeof src, "/tmp/wwflit_%d_ovf.ww", getpid());
|
||||||
|
if (writesrc(src, overflow_src) != 0) return 1;
|
||||||
|
snprintf(cmd, sizeof cmd, "%s %s >/dev/null 2>&1", w6c, src);
|
||||||
|
if (runwait(cmd) == 0) {
|
||||||
|
fprintf(stderr, "floatlit: w6c ACCEPTED overflow literal\n");
|
||||||
|
fail++;
|
||||||
|
}
|
||||||
|
snprintf(cmd, sizeof cmd, "%s %s >/dev/null 2>&1", w6c_ww, src);
|
||||||
|
if (runwait(cmd) == 0) {
|
||||||
|
fprintf(stderr, "floatlit: w6c_ww ACCEPTED overflow literal\n");
|
||||||
|
fail++;
|
||||||
|
}
|
||||||
|
unlink(src);
|
||||||
|
|
||||||
|
if (fail) {
|
||||||
|
fprintf(stderr, "floatlit: %d check(s) failed\n", fail);
|
||||||
|
return 1;
|
||||||
|
}
|
||||||
|
printf("floatlit: vectors byte-id + runtime-correct, "
|
||||||
|
"overflow rejected by both stages\n");
|
||||||
|
return 0;
|
||||||
|
}
|
||||||
@@ -82,6 +82,10 @@ static const struct ent ents[] = {
|
|||||||
{ .fixture = "lib/regex/regex_test.ww", .mode = M_ID },
|
{ .fixture = "lib/regex/regex_test.ww", .mode = M_ID },
|
||||||
{ .fixture = "lib/strconv/test/decimaltest.ww", .mode = M_ID },
|
{ .fixture = "lib/strconv/test/decimaltest.ww", .mode = M_ID },
|
||||||
{ .fixture = "lib/strconv/test/ftostest.ww", .mode = M_ID },
|
{ .fixture = "lib/strconv/test/ftostest.ww", .mode = M_ID },
|
||||||
|
/* graduated from #59.10 DIVERGE by the #62 float-literal fold fix
|
||||||
|
* (wwstage lexer now folds through strconv.stof64, matching
|
||||||
|
* cstage's strtod bit-for-bit) */
|
||||||
|
{ .fixture = "lib/strconv/test/stoftest.ww", .mode = M_ID },
|
||||||
{ .fixture = "lib/strconv/test/inttest.ww", .mode = M_ID },
|
{ .fixture = "lib/strconv/test/inttest.ww", .mode = M_ID },
|
||||||
{ .fixture = "lib/strings/stringstest.ww", .mode = M_ID },
|
{ .fixture = "lib/strings/stringstest.ww", .mode = M_ID },
|
||||||
{ .fixture = "lib/temp/temptest.ww", .mode = M_ID },
|
{ .fixture = "lib/temp/temptest.ww", .mode = M_ID },
|
||||||
@@ -125,8 +129,7 @@ static const struct ent ents[] = {
|
|||||||
.mode = M_DIVERGE, .cite = "#59.8" },
|
.mode = M_DIVERGE, .cite = "#59.8" },
|
||||||
{ .fixture = "lib/os/stattest.ww",
|
{ .fixture = "lib/os/stattest.ww",
|
||||||
.mode = M_DIVERGE, .cite = "#59.9" },
|
.mode = M_DIVERGE, .cite = "#59.9" },
|
||||||
{ .fixture = "lib/strconv/test/stoftest.ww",
|
/* #59.10 stoftest graduated to M_ID above (#62 fix) */
|
||||||
.mode = M_DIVERGE, .cite = "#59.10" },
|
|
||||||
{ .fixture = "lib/ww/lex/toktest.ww",
|
{ .fixture = "lib/ww/lex/toktest.ww",
|
||||||
.mode = M_DIVERGE, .cite = "#59.11" },
|
.mode = M_DIVERGE, .cite = "#59.11" },
|
||||||
/* asttest resolves `import tok` via -I lib/ww/lex, cf 905 */
|
/* asttest resolves `import tok` via -I lib/ww/lex, cf 905 */
|
||||||
|
|||||||
Reference in New Issue
Block a user