This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [LTO merge][8/15] GIMPLE streamer
On Wed, 30 Sep 2009, Diego Novillo wrote:
> Bytecode portability is a good goal, but it was never a design
> parameter (at least not to us). To fix these issues, we will need to
> insert a translation layer in the streamer.
>
> Would you object if this work is done as a follow-up during stage3 or
> may be as a 4.6 feature? The majority of use cases for LTO will
> involve bytecode generation and consumption in the same build.
I think it's something reasonable to fix on trunk during stage 3. I'd
consider it more important than the misuses of assertions that could cause
ICEs on invalid streamed data (I'd expect it to be rather easier to fix
than making bytecode consumption robust against fuzzed data, as well as
more useful in practice).
At a minimum, issues from comments on patches that are not addressed for
the merge should have PRs filed in Bugzilla, and the copyright/license
notice issues with various patches and the areas they are known to cause
regressions will need fixing for the merge. Otherwise, it's up to
maintainers / reviewers in the various parts of the compiler what needs
fixing to merge each patch, and in that regard I am only reviewing patches
1, 5 and parts of 6.
--
Joseph S. Myers
joseph@codesourcery.com