[Bug c++/110468] New: internal compiler error in nothrow_spec_p (g++ > 11)

[Bug c++/110468] New: internal compiler error in nothrow_spec_p (g++ > 11)

[gococobu at 7art dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

            Bug ID: 110468
           Summary: internal compiler error in nothrow_spec_p (g++ > 11)
           Product: gcc
           Version: og12 (devel/omp/gcc-12)
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: gococobu at 7art dot org
  Target Milestone: ---

Created attachment 55420
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55420&action=edit
Command line, stack trace

Compiled without problems until g++ v12, v13.

Internal compiler error when using an alias of a boost::variant inside a struct
template.

Attached command line and stack trace (the problem is that it's not a small
codebase, can't attach the preprocessed code, it's 16M).

First time filing a bug, I hope I'm following the guidelines. I don't think
it's a duplicate of any I can see (although it might be connected).

[Bug c++/110468] internal compiler error in nothrow_spec_p (g++ > 11)

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2023-06-28
     Ever confirmed|0                           |1
                 CC|                            |mpolacek at gcc dot gnu.org
            Version|og12 (devel/omp/gcc-12)     |14.0
             Status|UNCONFIRMED                 |WAITING

--- Comment #1 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
Please gzip/xz the .ii file and attach it here.  Without it, we can't do
anything.

(You can try <https://gcc.gnu.org/bugs/minimize.html>.)

[Bug c++/110468] internal compiler error in nothrow_spec_p (g++ > 11)

[gococobu at 7art dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #2 from gococobu at 7art dot org ---
Thanks for your reply. Even compressed, it is too big for bugzilla (1.5M .xz,
2.1M .gzip)

Maybe you can close this bug, and I'll resubmit when/if I get a smaller
testcase for this specific bug? (right now, I'm just trying to go around it)

[Bug c++/110468] internal compiler error in nothrow_spec_p (g++ > 11)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to gococobu from comment #2)
> Thanks for your reply. Even compressed, it is too big for bugzilla (1.5M
> .xz, 2.1M .gzip)
> 
> Maybe you can close this bug, and I'll resubmit when/if I get a smaller
> testcase for this specific bug? (right now, I'm just trying to go around it)

Is there a place where you can place the compressed temporary and I can have a
go at reducing it.

[Bug c++/110468] internal compiler error in nothrow_spec_p (g++ > 11)

[gococobu at 7art dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #4 from gococobu at 7art dot org ---
Sure, I uploaded it to https://x0x.pw/gcc_bug_110468/

I think the code throwing the ICE is correct... but can't really attest to it.

Please tell me if you need something else.

[Bug c++/110468] internal compiler error in nothrow_spec_p (g++ > 11)

[gococobu at 7art dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #5 from gococobu at 7art dot org ---
(by the code being correct I mean in terms of standards, best practices, etc...
it definitely does compile in gcc++ 10.2 and 11.0)

[Bug c++/110468] internal compiler error in nothrow_spec_p (g++ > 11)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to gococobu from comment #4)
> Sure, I uploaded it to https://x0x.pw/gcc_bug_110468/
> 
> I think the code throwing the ICE is correct... but can't really attest to
> it.
> 
> Please tell me if you need something else.

Thanks, I can reproduce the ICE, I am reducing it now.

In file included from src/client/actions/action.cpp:166:
./src/include/params/parvar_import.hpp: At global scope:
./src/include/params/parvar_import.hpp:35:13: internal compiler error: in
nothrow_spec_p, at cp/except.cc:1199
0xea1209 nothrow_spec_p(tree_node const*)
        /home/apinski/src/upstream-gcc-git/gcc/gcc/cp/except.cc:1199
0xd00e2c set_flags_from_callee(tree_node*)
        /home/apinski/src/upstream-gcc-git/gcc/gcc/cp/call.cc:341

[Bug c++/110468] internal compiler error in nothrow_spec_p (g++ > 11)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|1                           |0
             Status|WAITING                     |UNCONFIRMED

[Bug c++/110468] internal compiler error in nothrow_spec_p (g++ > 11)

[gococobu at 7art dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #7 from gococobu at 7art dot org ---
Being a newbie, I don't want to add too much noise to this... but the ICE
doesn't appear when `multivar def` doesn't have any initializer. If assigned
(or with brace enclosed init), it does crash, even if done explicitly (i.e. ` =
boost::variant<boost::blank>()`).

[Bug c++/110468] internal compiler error in nothrow_spec_p (g++ > 11)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to gococobu from comment #7)
> Being a newbie, I don't want to add too much noise to this... but the ICE
> doesn't appear when `multivar def` doesn't have any initializer. If assigned
> (or with brace enclosed init), it does crash, even if done explicitly (i.e.
> ` = boost::variant<boost::blank>()`).

No that gives a hint of what might be going on internally. In fact the
backtrace inside the compiler does have references to nsdmi even.

[Bug c++/110468] [12/13/14 regression] Internal compiler error in nothrow_spec_p

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #9 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 55422
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55422&action=edit
Reduced testcase

Attached is the reduced testcase (which was modified back to be valid).

[Bug c++/110468] [12/13/14 regression] Internal compiler error in nothrow_spec_p

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |10.4.0, 11.4.0
   Target Milestone|---                         |12.4
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
   Last reconfirmed|2023-06-28 00:00:00         |2023-06-29
      Known to fail|                            |12.1.0, 12.3.0, 13.1.0,
                   |                            |14.0

--- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.

Note in the reduced testcase, m needs to be a template, variant needs to be a
template but not depedent and hh<T>::value needs to be a value depedent.
Also note it works with -std=c++11 .

Slightly more reduced, changing `hh<T>::value` to just `T>0` as mentioned it
just needs to be value depedent:
```
template <int T> struct variant {
  variant() noexcept(T>0) {}
};
template <int> 
struct m {
  variant<1> def{};
};
struct v {
  v(m<1>);
};
v t = {{}};
```

[Bug c++/110468] [12/13/14 regression] Internal compiler error in nothrow_spec_p

[gococobu at 7art dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #11 from gococobu at 7art dot org ---
Thank you very much for your work, Andrew. I'll remove the files from the
posted URL now.

[Bug c++/110468] [12/13/14 regression] Internal compiler error in nothrow_spec_p

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

Patrick Palka <ppalka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |ppalka at gcc dot gnu.org
                 CC|                            |ppalka at gcc dot gnu.org
             Status|NEW                         |ASSIGNED

--- Comment #12 from Patrick Palka <ppalka at gcc dot gnu.org> ---
Started with r12-3559-g22806064a67cf3

[Bug c++/110468] [12/13/14 regression] Internal compiler error in nothrow_spec_p

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #13 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Patrick Palka <ppalka@gcc.gnu.org>:

https://gcc.gnu.org/g:9479da4515f7d019b4ef282d0e21536431c44f71

commit r14-2199-g9479da4515f7d019b4ef282d0e21536431c44f71
Author: Patrick Palka <ppalka@redhat.com>
Date:   Thu Jun 29 16:10:18 2023 -0400

    c++: NSDMI instantiation during overload resolution [PR110468]

    Here we find ourselves instantiating the NSDMI for A<1>::m when
    computing argument conversions during overload resolution, and
    thus tf_conv is set.  The flag causes mark_used for the constructor
    used in the NSDMI to exit early and not instantiate its noexcept-spec,
    which eventually leads to an ICE from nothrow_spec_p.

    This patch fixes this by clearing any special tsubst flags during
    instantiation of an NSDMI, since the result should be independent of
    the context that requires the instantiation.

            PR c++/110468

    gcc/cp/ChangeLog:

            * init.cc (maybe_instantiate_nsdmi_init): Mask out all
            tsubst flags except for tf_warning_or_error.

    gcc/testsuite/ChangeLog:

            * g++.dg/cpp0x/noexcept79.C: New test.

[Bug c++/110468] [12/13/14 regression] Internal compiler error in nothrow_spec_p

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #14 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-13 branch has been updated by Patrick Palka
<ppalka@gcc.gnu.org>:

https://gcc.gnu.org/g:23c64450cbce916547142519a72075d1930c8282

commit r13-7517-g23c64450cbce916547142519a72075d1930c8282
Author: Patrick Palka <ppalka@redhat.com>
Date:   Thu Jun 29 16:10:18 2023 -0400

    c++: NSDMI instantiation during overload resolution [PR110468]

    Here we find ourselves instantiating the NSDMI for A<1>::m when
    computing argument conversions during overload resolution, and
    thus tf_conv is set.  The flag causes mark_used for the constructor
    used in the NSDMI to exit early and not instantiate its noexcept-spec,
    which eventually leads to an ICE from nothrow_spec_p.

    This patch fixes this by clearing any special tsubst flags during
    instantiation of an NSDMI, since the result should be independent of
    the context that requires the instantiation.

            PR c++/110468

    gcc/cp/ChangeLog:

            * init.cc (maybe_instantiate_nsdmi_init): Mask out all
            tsubst flags except for tf_warning_or_error.

    gcc/testsuite/ChangeLog:

            * g++.dg/cpp0x/noexcept79.C: New test.

    (cherry picked from commit 9479da4515f7d019b4ef282d0e21536431c44f71)

[Bug c++/110468] [12/13/14 regression] Internal compiler error in nothrow_spec_p

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

--- Comment #15 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-12 branch has been updated by Patrick Palka
<ppalka@gcc.gnu.org>:

https://gcc.gnu.org/g:dfbe77ab8a70bc7eb48a96775da04095d933a3a6

commit r12-9767-gdfbe77ab8a70bc7eb48a96775da04095d933a3a6
Author: Patrick Palka <ppalka@redhat.com>
Date:   Thu Jun 29 16:10:18 2023 -0400

    c++: NSDMI instantiation during overload resolution [PR110468]

    Here we find ourselves instantiating the NSDMI for A<1>::m when
    computing argument conversions during overload resolution, and
    thus tf_conv is set.  The flag causes mark_used for the constructor
    used in the NSDMI to exit early and not instantiate its noexcept-spec,
    which eventually leads to an ICE from nothrow_spec_p.

    This patch fixes this by clearing any special tsubst flags during
    instantiation of an NSDMI, since the result should be independent of
    the context that requires the instantiation.

            PR c++/110468

    gcc/cp/ChangeLog:

            * init.cc (maybe_instantiate_nsdmi_init): Mask out all
            tsubst flags except for tf_warning_or_error.

    gcc/testsuite/ChangeLog:

            * g++.dg/cpp0x/noexcept79.C: New test.

    (cherry picked from commit 9479da4515f7d019b4ef282d0e21536431c44f71)

[Bug c++/110468] [12/13/14 regression] Internal compiler error in nothrow_spec_p

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110468

Patrick Palka <ppalka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|ASSIGNED                    |RESOLVED

--- Comment #16 from Patrick Palka <ppalka at gcc dot gnu.org> ---
Fixed for GCC 12.4/13.2/14, thanks for the bug report!