https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90760
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jakub at gcc dot gnu.org
--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I wonder if we couldn't do:
--- gcc/symtab.c.jj 2019-06-06 11:06:36.034399596 +0200
+++ gcc/symtab.c 2019-06-10 19:21:39.273909981 +0200
@@ -1583,7 +1583,7 @@ symtab_node::set_section (symtab_node *n
void
symtab_node::set_section (const char *section)
{
- gcc_assert (!this->alias);
+ gcc_assert (!this->alias || !this->analyzed);
call_for_symbol_and_aliases
(symtab_node::set_section, const_cast<char *>(section), true);
}
i.e. allow setting the section name until the varpool node is actually
analyzed.
With that change
void bar (void) {}
void foo (void) __attribute__ ((alias ("bar")));
void foo (void) __attribute__ ((section ("baz")));
will no longer ICE, but will be rejected similarly to how
void bar (void) {}
void foo (void) __attribute__ ((alias ("bar"), section ("baz")));
is rejected (different section name for alias vs. what it aliases with).
Honza, your thoughts?
