Re: staging is broken (again) with --enable-debugging-features on x86_64

Etienne Gagnon wrote:
> Hi Chris,
> 
> Chris Pickett wrote:
> 
>> The problem is that the __clear_cache call is only included if the 
>> inlined threading engine is used.  IMO, the check for 
>> SABLEVM_INLINED_THREADED_INTERPRETER should be moved outside of 
>> _svmf_iflush() to wherever _svmf_iflush() is called -- what if 
>> somebody else besides the inlined engine ever wants to call it?
> 
> 
> Why would anyone call _svmf_iflush() if the incline engine is not used?
> 
> The goal is for the inline, direct, and switch interpreters to be 
> increasingly portable, switch being portable in minutes of work. Dealing 
> with instruction-cache flushing is kind of tricky, and we want to avoid 
> it in the switch engine.

I thought it might be useful elsewhere, at some point, because it's a 
general-purpose memory operation.  The call is already #ifdef'd away in 
prepare_code.c (see follow-up email) if you're worried about the other 
interpreters.

If you don't want that, then you could simply put a dummy statement that 
reads from the pword parameter in _svmf_iflush(), to prevent errors 
about pword not being used.

Cheers,
Chris