From: Richard Biener <richard.guenther@gmail.com>
To: GCC Patches <gcc-patches@gcc.gnu.org>, Jan Hubicka <jh@suse.cz>
Subject: Re: [PR 70965] Schedule extra pass_rebuild_cgraph_edges
Date: Fri, 25 Nov 2016 08:48:00 -0000 [thread overview]
Message-ID: <CAFiYyc0asNFYVn2nYLOBPbqda2HGF379VUM1oLkyiZL7ujfhkg@mail.gmail.com> (raw)
In-Reply-To: <20161124164456.m7aavu3k76zpc7z5@virgil.suse.cz>
On Thu, Nov 24, 2016 at 5:44 PM, Martin Jambor <mjambor@suse.cz> wrote:
> Hi,
>
> after discussing this with Honza, we have decided that scheduling an
> extra pass_rebuild_cgraph_edges after pass_fixup_cfg is the correct
> way to keep the cgraph consistent with gimple IL when early IPA passes
> need it, such as is the case with the testcase in PR 70965.
>
> While needing an extra pass is never nice, this is a consequence of
> splitting pass_build_ssa_passes out of early optimization passes so
> that pass_chkp can be in between.
>
> The patch below fices the ICE in PR 70965 and has passed bootstrap and
> testing on x86_64-linux. OK for trunk?
Ok.
Richard.
> Thanks,
>
> Martin
>
>
> 2016-11-24 Martin Jambor <mjambor@suse.cz>
>
> gcc/
> * passes.def (pass_build_ssa_passes): Add pass_rebuild_cgraph_edges.
>
> gcc/testsuite/
> * g++.dg/pr70965.C: New test.
> ---
> gcc/passes.def | 1 +
> gcc/testsuite/g++.dg/pr70965.C | 21 +++++++++++++++++++++
> 2 files changed, 22 insertions(+)
> create mode 100644 gcc/testsuite/g++.dg/pr70965.C
>
> diff --git a/gcc/passes.def b/gcc/passes.def
> index 85a5af0..5faf17f 100644
> --- a/gcc/passes.def
> +++ b/gcc/passes.def
> @@ -56,6 +56,7 @@ along with GCC; see the file COPYING3. If not see
> NEXT_PASS (pass_build_ssa_passes);
> PUSH_INSERT_PASSES_WITHIN (pass_build_ssa_passes)
> NEXT_PASS (pass_fixup_cfg);
> + NEXT_PASS (pass_rebuild_cgraph_edges);
> NEXT_PASS (pass_build_ssa);
> NEXT_PASS (pass_warn_nonnull_compare);
> NEXT_PASS (pass_ubsan);
> diff --git a/gcc/testsuite/g++.dg/pr70965.C b/gcc/testsuite/g++.dg/pr70965.C
> new file mode 100644
> index 0000000..d8a2c35
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/pr70965.C
> @@ -0,0 +1,21 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -std=c++11" } */
> +
> +struct A {};
> +struct B {};
> +struct C { using p = int *; template <typename> using ra = A; };
> +struct J : C { template <typename> struct K { typedef C::ra<int> o; }; };
> +template <typename> struct D
> +{
> + struct H : J::K<int>::o { H (J::p, A) : J::K<int>::o () {} };
> + H d;
> + D (const char *, const A &x = A ()) : d (0, x) {}
> +};
> +extern template class D<char>;
> +enum L { M };
> +struct F { virtual char *foo (); };
> +template <class> struct I : B { static int foo (int) {} };
> +struct G { typedef I<int> t; };
> +void foo (int) { G::t::foo (0); }
> +void bar (const D<char> &, const D<int> &, int, L);
> +void baz () try { foo (0); } catch (F &e) { bar (e.foo (), "", 0, M); }
> --
> 2.10.2
>
next prev parent reply other threads:[~2016-11-25 8:48 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-11-24 16:45 Martin Jambor
2016-11-25 8:48 ` Richard Biener [this message]
2016-11-25 13:11 ` Jan Hubicka
2016-12-02 15:48 ` Martin Jambor
2016-12-02 16:15 ` Jan Hubicka
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAFiYyc0asNFYVn2nYLOBPbqda2HGF379VUM1oLkyiZL7ujfhkg@mail.gmail.com \
--to=richard.guenther@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=jh@suse.cz \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).