2008/7/10 Bill Maddox <maddox@???>:
> This patch avoids the failure in Rafael's test case, and serializes
> TREE_BINFO as
> intended in a manner analogous to the treatment of types and declarations.
>
> * lto-function-out.c (lto_static_init):
> Back out reset of lto_types_needed_for bit for
> TREE_BINFO nodes.
> * lto-function-in.c (input_binfo): Restore reading
> of flags and check for presence of line info.
> (input_tree_operand): Add test for tree code TREE_BINFO
> where we check for special handling of type and decl nodes.
>
> I am running the regression tests on it now.
Should be fine if bootstrap and all regression tests are. We were
reading things twice because binfo was passing the test
if (TREE_CODE_CLASS (code) != tcc_type && TREE_CODE_CLASS (code) !=
tcc_declaration)
I was not sure which one to cut out :-)
> --Bill
>
Thanks,
--
Rafael Avila de Espindola
Google Ireland Ltd.
Gordon House
Barrow Street
Dublin 4
Ireland
Registered in Dublin, Ireland
Registration Number: 368047
