Hi!
Perhaps the reason for this issue is that QuickJS does not meet the ES
standard，but I have some questions listed below:
Q1: Is it a bug of QuickJS?
Q2: if Q1 is yes, will this bug be fixed in the future, or has it been fixed?
Looking forward to your replay, thank you very much.
Cheers,
Houyou Yao
------------------ Original ------------------
From: "Mario Gliewe";<mag@xxxxxxxxx>;
Send time: Saturday, Sep 7, 2019 4:12 AM
To: "quickjs-devel"<quickjs-devel@xxxxxxxxxxxxx>;
Subject: [quickjs-devel] Re: A possible issue about the scope of variablein
QuickJS
Looks like what's missing is 6.a.: SameValue(func, %eval%) == true during
execution of bytecode.
https://tc39.es/ecma262/#sec-function-calls-runtime-semantics-evaluation
regards,
o.
On Fri, Sep 06, 2019 at 05:59:39PM +0200, quickjs mailing list wrote:
Hi,Looks like what's missing is 6.a.: SameValue(func, %eval%) == true during
On Fri, Sep 06, 2019 at 04:22:40PM +0200, Mario Gliewe wrote:
very funny, seems to be specific to the eval() function..It's due to special handling in js_parse_postfix_expr. If it's named eval,
i tried the following, which passed ok:
--snip--
let evil=666;
function test1(evil) {
console.log('the evil is ',evil);
eval('console.log("the evil is ",evil)');
}
function test2() {
test1(42);
};
test2();
function test3(Object) {
console.log('Object is ', Object);
}
function test4() {
console.log('Object is ', Object);
}
function test5() {
test3(' void');
};
test4();
test5();
--snap--
it will do eval, no matter what. And this is decided at parse time and not
during execution:
execution of bytecode.
https://tc39.es/ecma262/#sec-function-calls-runtime-semantics-evaluation
regards,
o.
if (call_type == FUNC_CALL_NORMAL) {
parse_func_call:
switch(opcode = get_prev_opcode(fd)) {
case OP_get_field:
/* keep the object on the stack */
fd->byte_code.buf[fd->last_opcode_pos] = OP_get_field2;
break;
case OP_scope_get_private_field:
/* keep the object on the stack */
fd->byte_code.buf[fd->last_opcode_pos] =
OP_scope_get_private_field2;
break;
case OP_get_array_el:
/* keep the object on the stack */
fd->byte_code.buf[fd->last_opcode_pos] =
OP_get_array_el2;
break;
case OP_scope_get_var:
{
JSAtom name;
int scope;
name = get_u32(fd->byte_code.buf +
fd->last_opcode_pos + 1);
scope = get_u16(fd->byte_code.buf +
fd->last_opcode_pos + 5);
here >>> if (name == JS_ATOM_eval && call_type ==
FUNC_CALL_NORMAL) {
/* direct 'eval' */
JS_FreeAtom(s->ctx, name);
fd->byte_code.size = fd->last_opcode_pos;
fd->last_opcode_pos = -1;
opcode = OP_eval;
} else {
/* verify if function name resolves to a simple
get_loc/get_arg: a function call inside a
`with`
statement can resolve to a method call of the
`with` context object
regards,
o.
greets
maG
On 06.09.19 14:35, 姚厚友 wrote:
Dear Sir or Madam:
I have a doubt about QuickJS, the detailed description is as follows:
## version: quickjs-2019-08-18
## Testcase:
var NISLFuzzingFunc = function(eval) {
print(eval("print(false);"));
};
var NISLParameter1 = function() {
print("run to NISLParameter1");
return true;
};
NISLFuzzingFunc(NISLParameter1);
## Command:
./ quickjs-2019-08-18/qjs testcase.js
## Output:
false
undefined
## Expected output:
run to NISLParameter1
true
## Description:
In the testcase above, "eval"（Line 2） should be the formal
parameter "eval", but QuickJS treats it as a global object "eval".
According to the scope of variable, I think this may be a bug in QuickJS.
Do you think so? Looking forward to your reply.
your faithfully
houyou yao
