http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54511
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|ASSIGNED |NEW
CC| |jason at gcc dot gnu.org
AssignedTo|jakub at gcc dot gnu.org |unassigned at gcc dot
| |gnu.org
--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-09-07
09:05:18 UTC ---
Seems the bug is on the C++ FE side, we have different a VAR_DECLs. One is
created by finish_anon_union -> build_anon_union_vars during tsubst_expr, but
when tsubsting the a = 0; assignment that VAR_DECL is copied to a new VAR_DECL
by tsubst_decl.
Apparently build_anon_union_vars is called twice, once during parsing, once
during instantiation, but all the pushdecls the second time look really weird
and don't affect what VAR_DECLs the following code actually uses.
I would expect that during tsubst either build_anon_union_vars isn't called at
all and instead pt.c somehow adjusts when needed the VAR_DECLs that were
already created (after tsubst_decling them), or create a new set of vars, but
somehow arrange those to be registered as the local specializations of the old
vars.
Jason, could you please look at this? Thanks.
