Duplicate main and __original_main

[juj]

It looks like if one builds with a main that is of form int main(){...}, there will be this kind of stub

 (func $main (; 3 ;) (param $0 i32) (param $1 i32) (result i32)
  (call $__original_main)
 )
 (func $__original_main (; 5 ;) (result i32)
    ...

that will be generated, but not inlined away even in -O3. Is that something that would be possible to optimize away?

[sbc100]

We do have special handling for that case in binaryen. There is a pass called
InlineMainPass (one the command line: "inline-main") which does just this.

Maybe this stopping being run at some point... we should investigate.

However, I somewhat surprised that this pass is even needed? Should binaryen
be able to always inline in cases like this? i.e. can't we make a general pass that basically say:

"If a function has just a single caller, and the function is not exported, inline the function into its caller.. no matter how big the function is?" Is there any situation where that is not desirable? Or we could be more specific and add "if the calling function is trivial.. i.e. its a wrapper"

[sbc100]

Ah, yes it looks like this was removed in #11995. I guess the normal inlining passes at -O3 not removing this, contrary to what the comment that PR says?

[sbc100]

From my brief experiments it looks like this is getting correctly inlined at -O2 and above.. but adding -g seems to prevent the inlining.

This seems wrong. -g is not supposed to effect code generation of optimization at all.

[stale]

This issue has been automatically marked as stale because there has been no activity in the past year. It will be closed automatically if no further activity occurs in the next 30 days. Feel free to re-open at any time if this issue is still relevant.