On Thu, Jun 24, 2021 at 12:11:23PM +0200, Jakub Jelinek wrote: > Hi! > > The following testcase ICEs during error-recovery, as build_c_cast calls > note_integer_operands on error_mark_node and that wraps it into > C_MAYBE_CONST_EXPR which is unexpected and causes ICE later on. > Seems most other callers of note_integer_operands check early if something > is error_mark_node and return before calling note_integer_operands on it.
Yeah. > The following patch fixes it by not calling on error_mark_node, another > possibility would be to handle error_mark_node in note_integer_operands and > just return it. > > Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk? This looks OK to me, thanks. > 2021-06-24 Jakub Jelinek <ja...@redhat.com> > > PR c/101171 > * c-typeck.c (build_c_cast): Don't call note_integer_operands on > error_mark_node. > > * gcc.dg/pr101171.c: New test. > > --- gcc/c/c-typeck.c.jj 2021-06-23 13:33:00.375434219 +0200 > +++ gcc/c/c-typeck.c 2021-06-23 17:51:17.501401208 +0200 > @@ -6131,6 +6131,7 @@ build_c_cast (location_t loc, tree type, > return value reflects this. */ > if (int_operands > && INTEGRAL_TYPE_P (type) > + && value != error_mark_node > && !EXPR_INT_CONST_OPERANDS (value)) > value = note_integer_operands (value); > > --- gcc/testsuite/gcc.dg/pr101171.c.jj 2021-06-23 17:56:53.409896019 > +0200 > +++ gcc/testsuite/gcc.dg/pr101171.c 2021-06-23 17:56:28.668227851 +0200 > @@ -0,0 +1,13 @@ > +/* PR c/101171 */ > +/* { dg-do compile } */ > +/* { dg-options "" } */ > + > +extern void foo (void); > +int x = 0x1234; > + > +void > +bar (void) > +{ > + if (x != (sizeof ((enum T) 0x1234))) /* { dg-error "conversion to > incomplete type" } */ > + foo (); > +} > > Jakub > Marek