Fix pointer crashes and round-trip const void *.

[cycript.git] / todo.txt

lol
unicode identifier support (native and \u)
support unions (right now 0-1 fields parsed as struct)
look into what String is, and whether to bridge it
think about bridging NSNumber with Number prototype
some JS callbacks don't use exception pointers at all...
a newline needs to not be allowed after a unary *
finish implementing default xml namespace statement
encode newlines in history for later replay (psql uses ^A)

consider replacing regex literals with constructors
  https://bugzilla.mozilla.org/show_bug.cgi?id=98409
numerification needs to use specific precision values
  https://bugzilla.mozilla.org/show_bug.cgi?id=5856
consider a mode where unicode string content is saved
  https://bugzilla.mozilla.org/show_bug.cgi?id=274152

NSDictionaries that have NSNumber keys don't getProperty
errors in another process aren't displayed; to fix this, parse errors should get converted to exceptions and thrown
CYPoolTry/Catch now carefully save the exception after it /no longer needs the exception/... uhh... wtf?
throw CYJSError should probably be replaced with CYThrow() across the board
figure out what to do about global context refs: I really really want to retain the bastards
the concept of NULL pooling is entirely incorrect and sad... bad... evil... need to work on this... really
NSArray's .toString() and .toLocaleString() fail hard, as Array.prototype.to*String are Array-specific
applyOnMainThread, when done at console, loops the cyonifier
special work needs to be done to correctly handle the "arguments" symbol: Declare("arguments", ...Special)
at the Program level I seem to be eating away all of the var statements
function pointers are ?; note that blocks are currently block_P = '?'
I should probably attempt to use the auto_ flag somehow to not do contexts_ push when compiling
Object_callAsFunction_toCYON should be implemented

[NSString stringWithString:""] crashes, on linux, not on mac
GS #defines should be _finline

replace procmod g+s with gdb's macosx_get_task_for_pid_rights
non-local return prologue is not being Replace()d: multipass compiler!
interpretation of documentation comments should be compiler-only and off by default
don't ever generate $ CYWith, in particular for CYLet... use CYFunctionExpression

semi-colon insertion warnings should mark after the previous token, not on the current one: makes a /lot/ more sense
new for pools should take a reference, not a pointer

output errors during Trampoline to some kind of log file

there is a reinterpret_cast<> that I replaced with an old-style cast on the output of dladdr

why do I never deallocate ffi closures?

make a reference holder for namearrays
rename names to accumulator when it is a name accumulator, and retake names for subset
I believe calls to mutableCopy will leak memory
consider pointerTo instead of toPointer
store the last exception in a variable
should WebUndefined be @undefined?

if something is a function, it should be output differently
evaluate usages of CYTry in ObjectiveC/Library for CYObjectiveTry
verify name targets of "incorrect number of arguments to"
maybe support __objc_{yes,no} intrinisic primitives

support new objective-c subscript notation:
  id key = ...;
  id value = object[key];
  -> id value = [object objectForKeyedSubscript:key];
  object[key] = newValue;
  -> [object setObject:newValue forKeyedSubscript:key];
and:
  NSUInteger idx = ...;
  id value = object[idx];
  -> id value = [object objectAtIndexedSubscript:idx];
  object[idx] = newValue;
  -> [object setObject:newValue atIndexedSubscript:idx];

use JSObjectMakeError to build exception objects
replace cy$getProperty's return value with JSValueRef
flatten compound expressions made up of compound expressions

cy# {{5,6};6;{7;{8;5}}}
got: {5,6;6;{7;8,5}}
wanted: 5,6,6,7,8,5

cy# for each (var i = 9 in [1, 2]) system.print(i);
got: {i=9;(function(e,t){for(t in e){i=e[t];system.print(i)}})([1,2])}
wanted: i=9,function(e,t){for(t in e)i=e[t],system.print(i)}([1,2])

with is being translated in a manner that doesn't handle variables
cy# function b() { for each (var t in a) {} }
function b(){var e,t,n;with({t:a,n:undefined})for(n in t)e=t[n]}

[[self description] cy$toCYON:] seems like a poor implementation of toCYON
implement cy$toJSON:inContext: for NSDictionary and NSArray

the non-local transform used for let is ludicrous... :( :(

cy# function q(e) { }
function q(e){var e}

CYCompound needs to be able to replace itself away without causing multiple Replace
fat arrows are currently output with too much verbosity

CYJSString <- wrap JavaScript strings and provide them back to Objective-C
ruby blocks need to return their last thing
Objective-C strings might should be cyonified using double-quotes

apparently you can have random escape sequences in strings, like \!
cycript -p with processes that have spaces doesn't work
CYDriver uses std::istream, but it should use std::streambuf
the mechanism I used to autodetect readline is horrible: use autoconf

constant flags on types are encoded using something horribly wrong
http://developer.apple.com/library/ios/#documentation/Cocoa/Conceptual/ObjCRuntimeGuide/Articles/ocrtTypeEncodings.html
http://gcc.gnu.org/onlinedocs/gcc/Type-encoding.html
http://gcc.gnu.org/onlinedocs/gcc/Legacy-type-encoding.html

using .withName from the grammar is a horrible hack that makes other uses impossible

blocks should be allowed to return blocks/functions: ModifiedType needs to go
  instead, TypedParameterList should be folded into a TypeModifier with the other types

LT_INIT([pic-only]) doesn't work, because the default is only used if unset, and it has a default setting...
even when using --pic-only, libtool insists on building every file twice... with the same fucking arguments