code wiki / _hdl_build / nx_js_unicode_probe.nx
`nx_js_unicode_probe.nx` source

1// nx_js_unicode_probe.nx -- ISOLATE the construct that blocks real bundles (2026-07-27).
2// netflix's 3.9MB bundle fails to parse at a lodash deburr map: {À:"A",Á:"A",...} -- object
3// keys that are NON-ASCII (UTF-8 multibyte) IDENTIFIERS, legal JS and near-universal in
4// bundled apps. This probe pins whether that is really the blocker, ASCII control first.
5// Each row prints its own verdict; exit 0 iff the ASCII controls pass (so a RED row is a
6// NAMED gap, not a broken probe).
7import "nx_syscalls.nx"
8import "nx_js_eval.nx"
9
10func up_p(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } sys_write(1, s, n); return 0 }
11func up_len(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } return n }
12
13// run a snippet; print PARSE-OK/PARSE-FAIL. returns 1 if it parsed+ran, else 0.
14func up_try(label: *u8, src: *u8) -> i64 {
15    let out: *i64 = sys_mmap(16) as *i64
16    let rc: i64 = js_run_source(src, up_len(src), out)
17    up_p(label)
18    if rc == 0 { up_p("  OK\x0a\x00" as *u8); return 1 }
19    up_p("  FAIL\x0a\x00" as *u8)
20    return 0
21}
22
23func main() -> i64 {
24    // --- ASCII controls: these MUST pass, else the probe itself is wrong.
25    let c1: i64 = up_try("C1 ascii ident key      {a:1}          \x00" as *u8,
26                         "var o={a:1}; o.a;\x00" as *u8)
27    let c2: i64 = up_try("C2 quoted non-ascii key {\"A-grave\":1}  \x00" as *u8,
28                         "var o={\"\xC3\x80\":1}; o[\"\xC3\x80\"];\x00" as *u8)
29    // --- the suspect: BARE non-ASCII identifier as an object key (lodash deburr shape).
30    let s1: i64 = up_try("S1 BARE non-ascii key   {A-grave:\"A\"}  \x00" as *u8,
31                         "var o={\xC3\x80:\"A\"}; 1;\x00" as *u8)
32    // --- the same character as a plain variable name (is it the KEY position or the LEXER?).
33    let s2: i64 = up_try("S2 non-ascii var name   var A-grave=1   \x00" as *u8,
34                         "var \xC3\x80=1; \xC3\x80;\x00" as *u8)
35    // --- the exact netflix shape: several such keys in one literal.
36    let s3: i64 = up_try("S3 netflix deburr shape {A-g:..,A-a:..} \x00" as *u8,
37                         "var m={\xC3\x80:\"A\",\xC3\x81:\"A\",\xC3\xA0:\"a\"}; 1;\x00" as *u8)
38
39    up_p("\x0aREAD: C1/C2 are controls (must be OK). S1-S3 name the gap.\x0a\x00" as *u8)
40    if c1 == 1 { if c2 == 1 {
41        up_p("controls GREEN -- any S FAIL above is a REAL, NAMED engine gap\x0a\x00" as *u8)
42        if s1 == 1 { if s2 == 1 { if s3 == 1 { up_p("verdict=NO-GAP (unicode idents already work)\x0a\x00" as *u8); return 0 } } }
43        up_p("verdict=GAP-CONFIRMED: non-ASCII identifiers\x0a\x00" as *u8)
44        return 0
45    } }
46    up_p("verdict=PROBE-BROKEN (a control failed)\x0a\x00" as *u8)
47    return 1
48}
nx_js_unicode_probe.nx source

`nx_js_unicode_probe.nx` source
