4 Pinned and volatile require type modifiers that can not be encoded
7 Properties and 17.6.3: Finish it.
9 Implement base indexer access.
13 These guys should only appear as part of an Invocation, so we
14 probably can afford to have a special callback:
16 Expression.ResolveAllowMemberGroups
18 This is only called by Invocation (or anyone that consumes
21 And the regular DoResolve and DoResolveLValue do emit the error
22 654 `Method referenced without argument list'.
24 Otherwise, a resolution will return a MethodGroupExpr which is
25 not guaranteed to have set its `Expression.Type' to a non-null
29 Needs an extra argument: `set' or `read', this will help
30 to track usage errors in variables (since AddressOf does not
31 really know what is it being used for, it could be either)
34 Pointer arithmetic operators (++ and -- are done)
35 Test for ++ and -- for non-builtin-types
38 readonly variables and ref/out
43 * Check for Final when overriding, if the parent is Final, then we cant
46 * Add test case for destructors
48 * Do not declare a .property on things that are just implementations, that
49 comes from the parent, just do the method.
51 * Currently the code path for 108/109 reporting is not being ran for methods
52 as we need to compare method signatures. But since we retrieve the expensive
53 method arguments in the method, we probably should do 108/109 processing there.
55 * Emit warning on hiding members without NEW not only in members.
57 * Implement visibility.
61 We do add variables in a number of places, and this is erroneous:
78 I am not reporting errors on visibility yet.
82 I have not figured out why the Microsoft version puts an
83 `instance' attribute, and I am not generating this `instance' attribute.
85 Explanation: The reason for the `instance' attribute on
86 indexers is that indexers only apply to instances
90 We need to make sure at *compile time* that the arguments in
91 the expression list of an array creation are always positive.
93 * Implement dead code elimination in statement.cs
95 It is pretty simple to implement dead code elimination in
100 the following wont work:
104 if x has indexers, the value of x [N] set is set to void. This needs to be
109 Multi-dim arrays are declared as [,] instead of [0..,0..]
111 * Break/Continue statements
113 A finally block should reset the InLoop/LoopBegin/LoopEnd, as
114 they are logically outside the scope of the loop.
116 * Break/continue part 2.
118 They should transfer control to the finally block if inside a try/catch
121 * Method Registration and error CS111
123 The way we use the method registration to signal 111 is wrong.
125 Method registration should only be used to register methodbuilders,
126 we need an alternate method of checking for duplicates.
130 extern string Property {
134 The get there should only have a semicolon
137 > // CSC sets beforefieldinit
139 > // .cctor will be generated by compiler
140 > public static readonly object O = new System.Object ();
141 > public static void Main () {}
150 Primary-expression, as it has now been split into
151 non-array-creation-expression and array-creation-expression.
153 * Static flow analysis
155 Required to warn about reachability of code and definite
156 assignemt as well as missing returns on functions.
160 The information when registering a method in InternalParameters
161 is duplicated, you can always get the types from the InternalParameters
163 * Emit modreq for volatiles
165 Handle modreq from public apis.
170 * Use of local temporary in UnaryMutator
172 We should get rid of the Localtemporary there for some cases
176 Do we really need to instanciate this variable all the time?
178 It could be static for all we care, and just use it for making
179 sure that there are no recursive invocations on it.
183 Since AppDomain exists, maybe we can get rid of all the stuff
184 that is part of the `compiler instance' and just use globals
190 Currently it calls the parent constructor before initializing fields.
191 It should do it the other way around.
193 * Use of EmitBranchable
195 Currently I use brfalse/brtrue in the code for statements, instead of
196 using the EmitBranchable function that lives in Binary
200 Currently ConvertImplicit will not catch things like:
202 - IntLiteral in a float context to generate a -FloatLiteral.
203 Instead it will perform an integer load followed by a conversion.
207 Write tests for the various reference conversions. We have
208 test for all the numeric conversions.
210 * Remove the tree dumper, cleanup `public readonly'
212 And make all the stuff which is `public readonly' be private unless
217 In Indexers and Properties, probably support an EmitWithDup
218 That emits the code to call Get and then leaves a this pointer
219 in the stack, so that later a Store can be emitted using that
220 this pointer (consider Property++ or Indexer++)
223 * Optimizations: variable allocation.
225 When local variables of a type are required, we should request
226 the variable and later release it when we are done, so that
227 the same local variable slot can be reused later on.
229 * Add a cache for the various GetArrayMethod operations.
233 Do not use StatementCollections, use ArrayLists of Statements,
234 and then "copy" it out to arrays. Then reuse the existing
235 Statement structures.
237 * TypeManager.FindMembers:
239 Instead of having hundreds of builder_to_blah hash table, have
240 a single one that maps a TypeBuilder `t' to a set of classes
241 that implement an interface that supports FindMembers.
246 * Use of lexer.Location in the parser
250 TOKEN nt TERMINAL nt TERMINAL nt3 {
251 $$ = new Blah ($2, $4, $6, lexer.Location);
254 This is bad, because the lexer.Location is for the last item in `nt3'
256 We need to change that to use this pattern:
258 TOKEN { oob_stack.Push (lexer.Location) } nt TERMINAL nt TERMINAL nt3 {
259 $$ = new Blah ($3, $5, $7, (Location) oob_stack.Pop ());
262 Notice how numbering of the arguments changes as the
263 { oob_stack.Push (lexer.Location) } takes a "slot" in the productions.
265 * local_variable_declaration
267 Not sure that this grammar is correct, we might have to
268 resolve this during semantic analysis.
273 Investigate what is the right value to return from `Emit' in
274 there (ie, for the `all code paths return')
279 Only create one `This' instance per class, and reuse it.
281 Maybe keep a pool of constants/literals (zero, 1)?
286 We would need to decode the shortname before we lookup members?