Issue 1403: Warn if int or uint is used in native fn declarations

[lht]

Please refer discussion of #1403.

Commit 37a2985 implements the check;
Commit 8926d5e fixes all occurrences in the current code base.

Is warn enough, or should we report errors?

I may not have used proper English for the warning message. Please help me correct that, if so. :)

[brson]

I'm ok with taking this as-is for now, but also a little wary of sprinkling little chunks of lint-style analysis all over the compiler.

[lht]

Thanks Brian! I'll rebase & merge this in a day or two if there was no objection raised.

[graydon]

I wonder how much this patch is, in reality, an argument for adopting LP64-style definition of int, to match C as much as possible by platform.

[lht]

This patch only uses types defined in ctypes. I think it should be the job of ctypes to decide which data model to use.
I should remove all the assumptions of data width in number initializations.

[nikomatsakis]

@graydon yes, I was about to pull this patch, but then I was wondering the same... maybe it would be better to just make "int and uint" be defined as "whatever type C int and unsigned are"? If we don't do that, however, I think we definitely need to make some kind of warnings because it will be a common misunderstanding.

[lht]

Rebase and committed as:

f03eb96 rustc: Warn when int or uint is used in a native type decl
dde4186 Use ctypes in native function declarations
04882d7 rustc: Use integer from ctypes consistently
7b1724f libstd: Revert an integer type casting

Maybe we need to file another issue for discussing adopting LP64?