awakening noop

[Stefan Reinauer]

Haven't seen these in a while, but they are back with the latest romcc
release:

apic.h:100.1: auto.c:123.19: auto.c:197.30: 0x735c00 noop Internal compiler warning: awakening noop?
apic.h:100.1: auto.c:129.19: auto.c:197.30: 0x736b90 noop Internal compiler warning: awakening noop?

Don't think they're saying anything else but that a dead subtree was
detected in the code?

Stefan

Stefan Reinauer  writes:
> Haven't seen these in a while, but they are back with the latest romcc
> release:
>
> apic.h:100.1: auto.c:123.19: auto.c:197.30: 0x735c00 noop Internal compiler
> warning: awakening noop?
>
> apic.h:100.1: auto.c:129.19: auto.c:197.30: 0x736b90 noop Internal compiler
> warning: awakening noop?
>
> Don't think they're saying anything else but that a dead subtree was
> detected in the code?

Pretty much.

These are actually comming from a slightly different path than
the previous ones and I don't think these are an error.  And
it has never been an error on the face of it.  However something
about the root cause last time was actually an error elsewhere
in the compiler and when I fixed that I managed to get -O2 to work.

So I decided to warn about them because I'm paranoid.

Eric

Stefan Reinauer  writes:
> Haven't seen these in a while, but they are back with the latest romcc
> release:
>
> apic.h:100.1: auto.c:123.19: auto.c:197.30: 0x735c00 noop Internal compiler
> warning: awakening noop?
>
> apic.h:100.1: auto.c:129.19: auto.c:197.30: 0x736b90 noop Internal compiler
> warning: awakening noop?
>
> Don't think they're saying anything else but that a dead subtree was
> detected in the code?

To clarify a little bit.

For the compiler to awakening a totally dead basic block/subtree is a
bug and will actually generate errors.

In this specific case the tree is only half dead which is very
peculiar.

Eric