https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106654

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |jakub at gcc dot gnu.org
             Status|NEW                         |ASSIGNED

--- Comment #12 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 53600
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=53600&action=edit
gcc13-pr106654.patch

Untested patch, with the more difficult cases optimized away during
gimplification, so only the most simple assertions are transformed into
if (cond); else __builtin_unreachable (); for now, but the FE handling should
be
all there, for C++ both as the standard attribute and [[gnu::assume (cond)]] or
__attribute__((assume (cond))) with the same behavior, for C just the latter
two.  Similarly, for C++ constant evaluation, it will diagnose only the simple
cases (mainly because we'd need to undo all changes to ctx->globals done during
evaluation of the assumption).  I believe such behavior in both places is
standard conforming, but obviously want to work especially on the former
(middle-end representation of those) soon.

Reply via email to