[Bug tree-optimization/119492] missing tail call due to address of the argument taken but NOT escaping

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119492

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #6)
> (In reply to Andrew Pinski from comment #5)
> > So looking into this further, I think PR 15387 got fixed correctly via PR
> > 49519 (and PR51323 and PR 50074). 
> 
> Note both mention not doing TER into the call to fix the issue:
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=49519#c18
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=19616#c11
> 
> I am still wondering if helps anyways since the testcase in comment #4 with
> the musttail uncomment works iff not TERing into the tail call.  
> 
> Oh wait https://gcc.gnu.org/bugzilla/show_bug.cgi?id=50074#c14 which takes
> about this exact loop and why it would be ok to look only at still
> internal_arg_pointer .
> 
> I am still thinking about not doing TER for tail calls might be the best
> option anyways. Since in the long run we want to get rid of TER and such.

Note thinking about this, TER for musttail calls should be disabled after
https://gcc.gnu.org/pipermail/gcc-patches/2025-March/679182.html due to this
too.