Resurrecting Stackmap insertion change

[nmdis1999]

These changes make sure that stackmap is not inserted after unmappable functions except control point and yk_promote calls. In previous PR #114 the stackmaps insertion was skipped as control point and yk_promote were treated as external functions, which resulted in some tests failing. This commit make sures that this problem don't occur while retaining the logic of previous PR.

[nmdis1999]

@ltratt @ptersilie the PR is ready for review.

[ptersilie]

I still have a feeling we might get away with omitting stackmaps for the control point too, but we can investigate that in a separate PR. It will require some changes to yk too.

[ltratt]

Which yklua tests has this been confirmed as working on?

[ltratt]

@nmdis1999 Ping on #119 (comment).

[nmdis1999]

@ltratt yes, I was figuring out why all promote tests were failing, and looks like the stackmap is not inserting after __yk_promote* calls because typo in my code. I was checking if the call starts with yk_promote instead of __yk_promote.
I have fixed this, just testing again and then will push in couple of minutes.

[nmdis1999]

@ltratt I have tested this on yk and all tests are passing, on yk lua all tests which are passing without this change also pass with my changes.

[ltratt]

Can we list which lua tests do and don't pass please?

[nmdis1999]

@ltratt These are the tests that pass and fail with my changes:

OK:  api.lua bwcoercion.lua calls.lua closure.lua code.lua constructs.lua coroutine.lua db.lua errors.lua events.lua files.lua gc.lua gengc.lua goto.lua literals.lua math.lua nextvar.lua pm.lua sort.lua strings.lua tpack.lua tracegc.lua utf8.lua vararg.lua
FAIL:  all.lua attrib.lua big.lua bitwise.lua cstack.lua heavy.lua locals.lua main.lua verybig.lua

which is same as without my changes. I have tested this with both my changes and without my changes and getting diff.

[ltratt]

Thanks! I think this is then ready for merge? @ptersilie OK?

[ptersilie]

Let's go.