Fixed implicit casts from size_t to int.

[neonichu]

Mostly by actually using size_t, except for the cases where negative values were considered valid, there I have stuck with long. The latter is of course all potential bugs, but at least they are more explicit now.

This makes it nicer to use cmark on iOS where -Wshorten-64-to-32 is a default warning.

[jgm]

@nwellnhof, I'd be interested in your feedback on the proposed change.

@neonichu, this seems incomplete - note the warnings on the travis build:
https://travis-ci.org/jgm/cmark/builds/48927694

[neonichu]

@jgm oh, it indeed is. It appears I didn't have -Wsign-compare set, will amend.

[nwellnhof]

I would prefer a patch that doesn't change the types of struct members or function parameters unless there's a really good reason. Using size_t for things like header levels just to fix some warnings isn't the right approach.

Instead, I'd propose to add explicit casts in the places where the warnings appear, and maybe add a TODO comment if the cast looks unsafe. Changing the types of local variables is OK.

[jgm]

+++ Nick Wellnhofer [Jan 30 15 13:43 ]:

I would prefer a patch that doesn't change the types of struct members or function parameters unless there's a really good reason. Using size_t for things like header levels just to fix some warnings isn't the right approach.

Instead, I'd propose to add explicit casts in the places where the warnings appear, and maybe add a TODO comment if the cast looks unsafe. Changing the types of local variables is OK.

This sounds like the right approach to me, too.

[nwellnhof]

This should be fixed now.