Re: Fix problem with late insns in sibcall

[Olivier Hainque <hainque at act-europe dot fr>]

Richard Kenner wrote:
> The confusing part to me is that after this discussion, the fix for the
> 3.2-based GNAT was to disable sibcall for PA due to a known bug (fixed
> for 3.3).  Then this showed up on another target.
> 
> Olivier, can you explain more?

 This history is correct. To add to the confusion, the problem occured with
 a sibcall to a nested function, which were later identified as potentially
 troublesome, as recently discussed.

 I tried to exhibit the problem again on a slightly modified case with a
 recent 3.2 based compiler, and could not. I'm also tempted to think that
 a fix of this kind would be better located in sibcall.c anyway.

 We could then decide to:

 1/ Leave things in their current state

 2/ Just revert the patch, considering that the testcase it fixed involved
    a sibcall to nested that we now avoid, and that other changes have occured
    since the initial fix, which possibly make it OBE.

 3/ Investigate further, which I could volonteer for :)


 I don't like 1 much. I have no strong opinion for 2 or 3 (or both).

 What do you think ?

 Olivier