Benjamin Smedberg wrote:
> I have a helper function where I'd like to be able to pass in a decl or a
> type, and get back standard information about it.
>
> This involves a rather unpleasant series of statements:
>
> function allowedOnStack(t) {
> if (t.typedef) {
> // dispatch typedef code
> } else if (t.isPointer or type.isReference) {
> // dispatch pointer/reference code
> } else if (t.isArray) {
> // dispatch array code
> } else if (t.kind) {
> switch (t.kind) {
> case "class":
> case "struct":
> // dispatch struct/class code
> break;
> case "union":
> // dispatch union code
> break;
> }
> }
> else if (t.precision) {
> // this is a primitive
> }
> else if (t.name == 'void') {
> // this is the meta 'void' type
> }
> else if (t.parameters) {
> // this must be a function type... but maybe it's a function decl...
> // how do I tell the difference?
> }
> else {
> // this must be a decl, because it isn't anything else
> // this seems rather fragile
> }
> }
>
> What I would really like is to make every type or decl object have a .kind,
> e.g.
>
> switch (t.kind) {
> case "void":
> // ...
> case "typedef":
> // ...
> case "array":
> // ...
> case "class":
> case "struct":
> // ...
> case "union":
> // ...
> case "primitive":
> // ...
> case "function":
> // ...
> case "decl":
> // I have really thought about whether I want a single "decl" kind or
> // separate "member", "local", "functiondecl" and "global" kinds.
> };
>
> I think I can implement my proposal, but I'd like to make sure others think
> it's a good idea, and have any thoughts on how decls should be presented,
> especially the difference between a function type and a function decl.
>
I don't think it's right. Correct way would be to use prototypes and
instanceOf. File a bug if you agree.
Taras
_______________________________________________
Dev-static-analysis mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-static-analysis
