REPL doesn't work when CONFIG_ASAN is set #502
Description
./build/qjs QuickJS-ng - Type ".help" for help qjs > RangeError: Maximum call stack size exceeded at ucs_length (<null>:0:0) at readline_print_prompt (<null>:0:0) at readline_start (<null>:0:0) at cmd_readline_start (<null>:0:0) at readline_handle_cmd (<null>:0:0) at handle_key (<null>:0:0) at handle_char (<null>:0:0) at handle_byte (<null>:0:0) at term_read_handler (<null>:0:0) I also needed this patch in order to make it that far, or ASAN trips:
diff --git a/quickjs.c b/quickjs.c index 22285d7..8d00d6f 100644 --- a/quickjs.c +++ b/quickjs.c @@ -6529,7 +6529,7 @@ static void build_backtrace(JSContext *ctx, JSValue error_obj, line_num1 = find_line_num(ctx, b, sf->cur_pc - b->byte_code_buf - 1, &col_num1); - atom_str = JS_AtomToCString(ctx, b->filename); + atom_str = b->filename ? JS_AtomToCString(ctx, b->filename) : NULL; dbuf_printf(&dbuf, " (%s", atom_str ? atom_str : "<null>"); JS_FreeCString(ctx, atom_str); if (line_num1 != -1) ASAN error:
================================================================= ==89091==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x000107900ab8 at pc 0x00010522ef0c bp 0x00016b4b79c0 sp 0x00016b4b7150 READ of size 1 at 0x000107900ab8 thread T0 #0 0x10522ef08 in printf_common(void*, char const*, char*)+0x84c (libclang_rt.asan_osx_dynamic.dylib:arm64e+0x22f08) #1 0x10522f2d8 in wrap_vsnprintf+0xac (libclang_rt.asan_osx_dynamic.dylib:arm64e+0x232d8) #2 0x1049126fc in dbuf_printf cutils.c:189 #3 0x104a2777c in build_backtrace quickjs.c:6533 #4 0x104a230c0 in JS_CallInternal quickjs.c:17111 #5 0x1049f1c4c in JS_CallInternal quickjs.c:15035 #6 0x1049f1c4c in JS_CallInternal quickjs.c:15035 #7 0x1049f1c4c in JS_CallInternal quickjs.c:15035 #8 0x1049f1c4c in JS_CallInternal quickjs.c:15035 #9 0x1049f1c4c in JS_CallInternal quickjs.c:15035 #10 0x1049f1c4c in JS_CallInternal quickjs.c:15035 #11 0x1049f1c4c in JS_CallInternal quickjs.c:15035 #12 0x1049f1c4c in JS_CallInternal quickjs.c:15035 #13 0x1049e2c08 in JS_Call quickjs.c:17159 #14 0x1048f2500 in call_handler quickjs-libc.c:2160 #15 0x1048f04f8 in js_os_poll quickjs-libc.c:2403 #16 0x1048dc55c in js_std_loop quickjs-libc.c:4020 #17 0x1048d2e58 in main qjs.c:526 #18 0x1a4b2bf24 (<unknown module>) 0x000107900ab8 is located 0 bytes after 40-byte region [0x000107900a90,0x000107900ab8) allocated by thread T0 here: #0 0x10525f244 in wrap_malloc+0x94 (libclang_rt.asan_osx_dynamic.dylib:arm64e+0x53244) #1 0x104a798b0 in js_def_malloc quickjs.c:1707 #2 0x1049748f8 in js_malloc_rt quickjs.c:1377 #3 0x104974a8c in js_mallocz_rt quickjs.c:1404 #4 0x104996ec4 in __JS_NewAtom quickjs.c:2748 #5 0x10499b134 in __JS_NewAtomInit quickjs.c:2848 #6 0x104976c30 in JS_InitAtoms quickjs.c:2600 #7 0x1049761c8 in JS_NewRuntime2 quickjs.c:1652 #8 0x10497d0a0 in JS_NewRuntime quickjs.c:1763 #9 0x1048d29c0 in main qjs.c:465 #10 0x1a4b2bf24 (<unknown module>) Not sure if legit bug or shenanigans...