main.ww was declaring `def PATH_MAX: u64 = 4096u64;` alongside the
imported os module's `export def PATH_MAX: i32 = 4096;` (lib/os/os.ww:91).
The duplicate triggered the cstage mod_map last-write-wins attribution
bug (#127): cstage emitted `main.PATH_MAX` for os's DATA entry too,
producing two 4096-byte slots and a wrong-module symbol. Both stages
shipped working binaries because LOADs all routed to main.PATH_MAX(SB),
but the divergence was real.
Removing the duplicate closes the only known #127 consumer per rule-7;
the defensive cstage cgen fix stays filed (task #11) until a new
duplicate-def collision surfaces. PATH_MAX use-sites in main.ww cast
`os.PATH_MAX: u64` at point-of-use for alloc/arithmetic, and use the
bare i32 form directly for slice .len assigns (which expect i32).
No cgen surface touched; byte-id-neutral on the 990-997 gates.
Hojun-Cho
90d31c5b41