Ref to function self should not create new object

[legendecas]

Picked from yodaos-project/ShadowNode#368

JerryScript-DCO-1.0-Signed-off-by: legendecas legendecas@gmail.com

[LaszloLango]

Please add a test for this as well.

[zherczeg]

I am also thinking about fixing this. I don't want to add a new argument for vm_run. We need to figure out something better.

[LaszloLango]

Measurement results (RPi2):

Benchmark	Heap (bytes)	Stack (bytes)	Perf (sec)
3d-cube.js	16376 -> 16376 : 0.000%	15888 -> 15896 : -0.050%	0.813 -> 0.814 : -0.046%
3d-raytrace.js	121600 -> 121600 : 0.000%	2840 -> 2840 : 0.000%	1.049 -> 1.053 : -0.398%
access-binary-trees.js	32752 -> 32752 : 0.000%	3304 -> 3304 : 0.000%	0.578 -> 0.576 : +0.390%
access-fannkuch.js	4968 -> 4968 : 0.000%	1584 -> 1592 : -0.505%	2.119 -> 2.124 : -0.249%
access-nbody.js	6248 -> 6248 : 0.000%	1560 -> 1568 : -0.513%	1.089 -> 1.096 : -0.642%
bitops-3bit-bits-in-byte.js	704 -> 704 : 0.000%	1528 -> 1536 : -0.524%	0.508 -> 0.510 : -0.365%
bitops-bits-in-byte.js	656 -> 656 : 0.000%	1528 -> 1536 : -0.524%	0.666 -> 0.665 : +0.207%
bitops-bitwise-and.js	496 -> 496 : 0.000%	1304 -> 1312 : -0.613%	0.925 -> 0.926 : -0.122%
bitops-nsieve-bits.js	98304 -> 98304 : 0.000%	1528 -> 1536 : -0.524%	1.142 -> 1.138 : +0.295%
controlflow-recursive.js	888 -> 888 : 0.000%	69784 -> 69784 : 0.000%	0.366 -> 0.366 : -0.063%
crypto-aes.js	33056 -> 33056 : 0.000%	2416 -> 2424 : -0.331%	0.883 -> 0.887 : -0.482%
crypto-md5.js	106488 -> 106488 : 0.000%	1924 -> 1924 : 0.000%	0.604 -> 0.604 : +0.073%
crypto-sha1.js	65528 -> 65528 : 0.000%	1640 -> 1648 : -0.488%	0.596 -> 0.598 : -0.322%
date-format-tofte.js	16376 -> 16376 : 0.000%	2056 -> 2072 : -0.778%	0.764 -> 0.754 : +1.255%
date-format-xparb.js	16376 -> 16376 : 0.000%	2496 -> 2512 : -0.641%	0.523 -> 0.524 : -0.210%
math-cordic.js	1776 -> 1776 : 0.000%	1528 -> 1536 : -0.524%	1.175 -> 1.184 : -0.778%
math-partial-sums.js	1232 -> 1232 : 0.000%	1488 -> 1496 : -0.538%	0.736 -> 0.737 : -0.145%
math-spectral-norm.js	6088 -> 6088 : 0.000%	1596 -> 1596 : 0.000%	0.539 -> 0.539 : -0.001%
string-base64.js	90792 -> 90792 : 0.000%	1560 -> 1568 : -0.513%	1.409 -> 1.400 : +0.634%
string-fasta.js	16376 -> 16376 : 0.000%	1512 -> 1520 : -0.529%	1.371 -> 1.373 : -0.138%
Geometric mean:	0.000%	-0.379%	-0.054%

Binary (bytes)	3f9dd0f	42ba64c	Gain
size	132664	132664	0.000%
.text	116540	116556	-0.014%
.rodata	11353	11353	0.000%
.bss	1575168	1575168	0.000%

[zherczeg]

I have bad news. We have to do this literally according to spec. See this example:

var f = function g() {
  console.log(g);
  eval("var g = 6");
  console.log(g);
}
f();

[legendecas]

@zherczeg: I have bad news. We have to do this literally according to spec. See this example:

var f = function g() {
  console.log(g);
  eval("var g = 6");
  console.log(g);
}
f();

Looks like the issue still exists with this PR. If we're going to address the issue in this PR?

[zherczeg]

Started to put together a correct solution #2634 . Not completed yet.

[akosthekiss]

Can we close this in favour of #2634 ? It seems to be close to completion.

[LaszloLango]

@akosthekiss yes we can close this, #2634 is the right fix for the original issue. This PR is just a workaround which does not work correctly on every testcase.