On Mon, 22 Aug 2022, Marek Polacek wrote:
> On Mon, Aug 22, 2022 at 10:16:23AM -0700, Andrew Pinski wrote:
> > On Mon, Aug 22, 2022 at 9:58 AM Marek Polacek via Gcc-patches
> > <gcc-patches@gcc.gnu.org> wrote:
> > >
> > > r11-4123 came without a test but I happened upon a nice test case that
> > > got fixed by that revision. So I think it'd be good to add it. The
> > > ICE was:
> > >
> > > phi-1.C: In constructor 'ElementManager::ElementManager()':
> > > phi-1.C:28:1: error: missing definition
> > > 28 | ElementManager::ElementManager() : array_(makeArray()) {}
> > > | ^~~~~~~~~~~~~~
> > > for SSA_NAME: _12 in statement:
> > > _10 = PHI <_12(3), _11(5)>
> > > PHI argument
> > > _12
> > > for PHI node
> > > _10 = PHI <_12(3), _11(5)>
> > > during GIMPLE pass: fixup_cfg
> > > phi-1.C:28:1: internal compiler error: verify_ssa failed
Interesting. I suspect the issue isn't actually in fixup_cfg though.
> > > Incidentally, do you think it's safe to backport r11-4123 to GCC 8
> > > (downstream)? Thanks.
It should be safe - the comment says it won't do any edge / PHI arg
re-ordering (which has proved to cause issues in the past).
> > > Tested on x86_64-pc-linux-gnu, ok for trunk?
> > >
> > > gcc/testsuite/ChangeLog:
> > >
> > > * g++.dg/tree-ssa/phi-1.C: New test.
> > > ---
> > > gcc/testsuite/g++.dg/tree-ssa/phi-1.C | 28 +++++++++++++++++++++++++++
> > > 1 file changed, 28 insertions(+)
> > > create mode 100644 gcc/testsuite/g++.dg/tree-ssa/phi-1.C
> > >
> > > diff --git a/gcc/testsuite/g++.dg/tree-ssa/phi-1.C
> > > b/gcc/testsuite/g++.dg/tree-ssa/phi-1.C
> > > new file mode 100644
> > > index 00000000000..26cade052e3
> > > --- /dev/null
> > > +++ b/gcc/testsuite/g++.dg/tree-ssa/phi-1.C
> > > @@ -0,0 +1,28 @@
> > > +// { dg-do compile { target c++11 } }
> > > +// { dg-options "-O --param early-inlining-insns=14" }
> >
> > It seems better to put it in g++.dg/torture and remove -O part.
> > And then it will be tested for more than before.
>
> Sure, I can move it.
The adjusted testcase is OK.
Thanks,
Richard.
> > Maybe even just mark makeArray as always_inline/inline and you might
> > be able to remove the param also.
>
> I can't, it doesn't reproduce the issue then.
>
> -- >8 --
> r11-4123 came without a test but I happened upon a nice test case that
> got fixed by that revision. So I think it'd be good to add it. The
> ICE was:
>
> phi-1.C: In constructor 'ElementManager::ElementManager()':
> phi-1.C:28:1: error: missing definition
> 28 | ElementManager::ElementManager() : array_(makeArray()) {}
> | ^~~~~~~~~~~~~~
> for SSA_NAME: _12 in statement:
> _10 = PHI <_12(3), _11(5)>
> PHI argument
> _12
> for PHI node
> _10 = PHI <_12(3), _11(5)>
> during GIMPLE pass: fixup_cfg
> phi-1.C:28:1: internal compiler error: verify_ssa failed
>
> gcc/testsuite/ChangeLog:
>
> * g++.dg/torture/phi-1.C: New test.
> ---
> gcc/testsuite/g++.dg/torture/phi-1.C | 28 ++++++++++++++++++++++++++++
> 1 file changed, 28 insertions(+)
> create mode 100644 gcc/testsuite/g++.dg/torture/phi-1.C
>
> diff --git a/gcc/testsuite/g++.dg/torture/phi-1.C
> b/gcc/testsuite/g++.dg/torture/phi-1.C
> new file mode 100644
> index 00000000000..69fb3d7ba38
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/torture/phi-1.C
> @@ -0,0 +1,28 @@
> +// { dg-do compile { target c++11 } }
> +// { dg-options "--param early-inlining-insns=14" }
> +
> +struct Element;
> +template <int _Nm> struct __array_traits { typedef Element _Type[_Nm]; };
> +template <int _Nm> struct array {
> + typename __array_traits<_Nm>::_Type _M_elems;
> +};
> +bool logLevel();
> +struct LogCapture {
> + void stream();
> +};
> +struct Element {
> + Element();
> + long data_;
> +};
> +using ElementArray = array<6>;
> +struct ElementManager {
> + ElementManager();
> + ElementArray array_;
> +};
> +static ElementArray makeArray() {
> + if (logLevel())
> + LogCapture().stream();
> + ElementArray foo;
> + return foo;
> +}
> +ElementManager::ElementManager() : array_(makeArray()) {}
>
> base-commit: b6316324fceaef431799a8b386de5cc9881d6898
>
--
Richard Biener <rguent...@suse.de>
SUSE Software Solutions Germany GmbH, Frankenstrasse 146, 90461 Nuernberg,
Germany; GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman;
HRB 36809 (AG Nuernberg)
