[PATCH] D25153: preprocessor supports `-dI` flag

[Steve O'Brien via cfe-commits]

elsteveogrande added a comment.

Ping again -- addressed all issues, looking ok now?

Note about this change + review process here.  First I know the every 1-2 day pinging generates noise on the lists, and for that  I'm SorryNotSorry :-p

I just want to ensure this is reaching at least the right people for feedback + eventually accepting.

Also this diff is pretty low-risk...

- It's short and doesn't change too much.
- What it does change is essentially opt-in with a new flag.  So this shouldn't break existing toolchains and projects, because nothing uses this yet.
- It's largely "additive" and again shouldn't change existing behavior, just allowing for new functionality.
- This new flag is for feature parity w/ gcc, and so there is some prescription for its behavior.  Though not officially documented (absent from GCC docs, for example, but with a small amount of info in `man 1 gcc`), it seems to work roughly the same.
- Has unit test coverage which I believe covers all possible usage cases (different ways to `#include` files).
- I stripped away unneeded code which was creating some sticking points, working around that for now, shortening this diff more.  (And reducing mystery and having to make decisions about e.g. string escaping and weird stuff.)
- Had quite a bit of feedback already from @rsmith, @majnemer, @vsk (thank you all!) -- addressed a number of issues and cleaned this up a lot.

Is this in acceptable shape?  Any objections to accept, land, then revert if catastrophe (doubt it for reasons above)?

Thanks again!


https://reviews.llvm.org/D25153