Multiple fixes #10
Multiple fixes #10
Conversation
Fixed keyed property usage
Fixed Construct with r0-r0 range usage.
String encoding fixer.
Fixed strings encoding output.
Replaced type checking for a reg value.
Simplify/fix_strings_encoding.py
Outdated
There was a problem hiding this comment.
Also tried something similar but I think this needs a v8 patch
https://github.com/v8/v8/blob/0e75d85d8e3467a536bca01d89d8a180a8bcffca/src/objects/string.cc#L725-L726
The issue is that v8 passes an utf-16 number to \\x%02x, so the output can be \x00 to \xffff
But when parsing we can't tell where the code point ends and the next one starts (e.g. \x7f34="缒" vs \x7f34="\x7f"+"34")
There was a problem hiding this comment.
According to JS syntax, there is only one interpretation of \xAB, and only one interpretation of \u{AABB} - so there are no \xAABB and no \u{AB}.
There was a problem hiding this comment.
Unfortunately the disassembler output doesn't adhere to js syntax and won't contain \u{AABB}
you can try compiling/decompiling the following script:
x = "\u{345}6"; // length 2
y = "\u{3456}"; // length 1 62 S> 0x1a1c74e41246 @ 0 : 13 00 LdaConstant [0]
64 E> 0x1a1c74e41248 @ 2 : 23 01 00 StaGlobal [1], [0]
78 S> 0x1a1c74e4124b @ 5 : 13 02 LdaConstant [2]
80 E> 0x1a1c74e4124d @ 7 : 23 03 02 StaGlobal [3], [2]
0x1a1c74e41250 @ 10 : 0e LdaUndefined
95 S> 0x1a1c74e41251 @ 11 : a9 Return
Constant pool (size = 4)
0x1a1c74e41259: [FixedArray] in OldSpace
- map: 0x06c319d80211 <Map(FIXED_ARRAY_TYPE)>
- length: 4
0: 0x1a1c74e41289 <String[2]: u#\x3456>
1: 0x06c319d838c1 <String[1]: #x>
2: 0x1a1c74e412a1 <String[1]: u#\x3456>
3: 0x06c319d838d9 <String[1]: #y>There was a problem hiding this comment.
I mean the disasm. Try to compile an umlaut and any unicode symbol, not its hex/u encoding.
There was a problem hiding this comment.
Still getting a similar result.. Could you please send an example how it looks for you?
There was a problem hiding this comment.
You right, There are three possible types of hex encoding:
There was a problem hiding this comment.
Ok, so what do you think we should do with that? Only fixing the disasm?
There was a problem hiding this comment.
Yeah it would be ideal if the disasm outputs valid utf-8/utf-16 if possible, or \u{HHHH} escapes which can be parsed with your python function
Unicode is pretty hard to get right (surrogate pairs, code points > 0xffff, etc).. could also merge the rest of this PR now and do it in another one if you want
Fixed keyed property usage.
Fixed Construct zero-range usage.
Fixing strings encoding.