[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 --- Comment #14 from Jonathan Wakely --- Commits r268728 and r269799 fixed this but had the wrong PR in the changelog.
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 Jan Hubicka changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED --- Comment #13 from Jan Hubicka --- Fixed.
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 Jason Merrill changed: What|Removed |Added Status|ASSIGNED|NEW Assignee|jason at gcc dot gnu.org |unassigned at gcc dot gnu.org --- Comment #12 from Jason Merrill --- (In reply to Jan Hubicka from comment #9) > This is caused by: > > _ZN4itpp1cILNS_1aE0EED2Ev/1 (itpp::c< >::~c() [with itpp::a > Referring: _ZN4itpp1cILNS_1aE0EED1Ev/2 (alias) > > this is a destructor which is pointed to from virutal table, but it has no > DECL_VIRTUAL_P flag set. This makes free_lang_data to modify its context and > point to outer namespace rather than the containing polymorphic type. > > Why we do not set VIRTUAL_P here? This is the base (D2) variant of the destructor, which is not virtual and does not appear in the vtable; the complete (D1) variant of the destructor is virtual and does appear in the vtable, and in this case is an alias for the base variant. I imagine that LTO is treating them as interchangeable, which causes trouble here. Unassigning myself.
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 Jason Merrill changed: What|Removed |Added Status|REOPENED|ASSIGNED CC||jason at gcc dot gnu.org Assignee|hubicka at gcc dot gnu.org |jason at gcc dot gnu.org
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 --- Comment #11 from Martin Liška --- Is Jason working on that? Can you please link a gcc-patches mailing list discussion (if there's any).
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 --- Comment #10 from Jan Hubicka --- Jason, I can avoid the ICE by simply not simplifying contexts for all C++ destructors. Index: tree.c === --- tree.c (revision 268728) +++ tree.c (working copy) @@ -5772,7 +5772,7 @@ free_lang_data_in_decl (tree decl, struc these are needed by devirtualization. */ if (TREE_CODE (decl) != FIELD_DECL && ((TREE_CODE (decl) != VAR_DECL && TREE_CODE (decl) != FUNCTION_DECL) - || !DECL_VIRTUAL_P (decl))) + || (!DECL_VIRTUAL_P (decl) && !DECL_CXX_DESTRUCTOR_P (decl DECL_CONTEXT (decl) = fld_decl_context (DECL_CONTEXT (decl)); } but it seems to me this is more a workaround - if the destructor is virtual, it should have VIRTUAL flag set. Or is there any reason we want to handle it differently then any other normal virutal function in the middle-end?
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 Jan Hubicka changed: What|Removed |Added CC||jason at redhat dot com --- Comment #9 from Jan Hubicka --- This is caused by: _ZN4itpp1cILNS_1aE0EED2Ev/1 (itpp::c< >::~c() [with itpp::a = itpp::b]) @0x770e9438 Type: function definition analyzed Visibility: forced_by_abi public weak comdat_group:_ZN4itpp1cILNS_1aE0EED5Ev one_only Same comdat group as: _ZN4itpp1cILNS_1aE0EED1Ev/2 Address is taken. Aux: @0x770e95a0 References: _ZTVN4itpp1cILNS_1aE0EEE/4 (addr)__gxx_personality_v0/8 (addr) Referring: _ZN4itpp1cILNS_1aE0EED1Ev/2 (alias) Function flags: body Called by: Calls: itpp::CFix::~CFix()/7 this is a destructor which is pointed to from virutal table, but it has no DECL_VIRTUAL_P flag set. This makes free_lang_data to modify its context and point to outer namespace rather than the containing polymorphic type. Why we do not set VIRTUAL_P here?
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089
Martin Liška changed:
What|Removed |Added
Priority|P3 |P1
--- Comment #8 from Martin Liška ---
It's still present for the following test-case:
$ cat 1.ii
namespace itpp {
template void b(a *c) { c[0].~a(); }
class CFix;
template class d {
void e(const char *);
CFix *data;
};
class CFix {
public:
virtual ~CFix();
};
template <> void d::e(const char *) { b(data); }
} // namespace itpp
$ cat 2.ii
namespace itpp {
enum a { b };
class CFix {
public:
virtual ~CFix();
};
template class c : CFix {
~c() {}
};
template class c<>;
} // namespace itpp
$ g++ -flto 1.ii 2.ii -shared -fPIC -O2
during IPA pass: devirt
lto1: internal compiler error: tree check: expected class ‘type’, have
‘declaration’ (namespace_decl) in type_with_linkage_p, at ipa-utils.h:185
0x6c7d51 tree_class_check_failed(tree_node const*, tree_code_class, char
const*, int, char const*)
/home/marxin/Programming/gcc/gcc/tree.c:9856
0xa3a7fd tree_class_check(tree_node const*, tree_code_class, char const*, int,
char const*)
/home/marxin/Programming/gcc/gcc/tree.h:3555
0xa3a7fd type_with_linkage_p(tree_node const*)
/home/marxin/Programming/gcc/gcc/ipa-utils.h:185
0xa3a7fd type_in_anonymous_namespace_p(tree_node const*)
/home/marxin/Programming/gcc/gcc/ipa-utils.h:221
0xa311ba maybe_record_node
/home/marxin/Programming/gcc/gcc/ipa-devirt.c:2531
0xa31c7c record_target_from_binfo
/home/marxin/Programming/gcc/gcc/ipa-devirt.c:2675
0xa31b4a record_target_from_binfo
/home/marxin/Programming/gcc/gcc/ipa-devirt.c:2687
0xa31fbd possible_polymorphic_call_targets_1
/home/marxin/Programming/gcc/gcc/ipa-devirt.c:2730
0xa37794 possible_polymorphic_call_targets(tree_node*, long,
ipa_polymorphic_call_context, bool*, void**, bool)
/home/marxin/Programming/gcc/gcc/ipa-devirt.c:3351
0xa39559 possible_polymorphic_call_targets(cgraph_edge*, bool*, void**, bool)
/home/marxin/Programming/gcc/gcc/ipa-utils.h:118
0xa39559 ipa_devirt
/home/marxin/Programming/gcc/gcc/ipa-devirt.c:3770
0xa39559 execute
/home/marxin/Programming/gcc/gcc/ipa-devirt.c:4088
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 --- Comment #7 from Dmitry G. Dyachenko --- (In reply to Dmitry G. Dyachenko from comment #6) > r267445 PASS sorry for noise 1. testcase from PR FAIL for me with r267445 2. real-world-code-compilation not FAIL
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 --- Comment #6 from Dmitry G. Dyachenko --- r267445 PASS
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089
Dmitry G. Dyachenko changed:
What|Removed |Added
Status|RESOLVED|REOPENED
Resolution|FIXED |---
--- Comment #5 from Dmitry G. Dyachenko ---
r267346 PASS
r267418 FAIL
PR87089]$ g++ -flto -fPIC -DPIC -shared a.o b.o
b.ii:6:8: warning: type ‘struct ASN1C_CertificateListAssertion’ violates the
C++ One Definition Rule [-Wodr]
6 | struct ASN1C_CertificateListAssertion : a{} * b;
|^
a.ii:2:8: note: a type with different bases is defined in another translation
unit
2 | struct ASN1C_CertificateListAssertion {
|^
during IPA pass: pure-const
lto1: internal compiler error: tree check: expected class ‘type’, have
‘declaration’ (namespace_decl) in type_with_linkage_p, at ipa-utils.h:185
0x732d05 tree_class_check_failed(tree_node const*, tree_code_class, char
const*, int, char const*)
/home/dimhen/src/gcc_current/gcc/tree.c:9856
0xc1880a maybe_record_node
/home/dimhen/src/gcc_current/gcc/ipa-devirt.c:2531
0xc1ec81 possible_polymorphic_call_targets(tree_node*, long,
ipa_polymorphic_call_context, bool*, void**, bool)
/home/dimhen/src/gcc_current/gcc/ipa-devirt.c:3340
0xc55506 possible_polymorphic_call_targets(cgraph_edge*, bool*, void**, bool)
/home/dimhen/src/gcc_current/gcc/ipa-utils.h:118
0xc55506 walk_polymorphic_call_targets
/home/dimhen/src/gcc_current/gcc/ipa.c:183
0xc57cd4 symbol_table::remove_unreachable_nodes(_IO_FILE*)
/home/dimhen/src/gcc_current/gcc/ipa.c:427
0x98d642 read_cgraph_and_symbols
/home/dimhen/src/gcc_current/gcc/lto/lto.c:2978
0x98d642 lto_main()
/home/dimhen/src/gcc_current/gcc/lto/lto.c:3362
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 Martin Liška changed: What|Removed |Added Status|ASSIGNED|RESOLVED Resolution|--- |FIXED --- Comment #4 from Martin Liška --- I can confirm it's fixed.
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 --- Comment #3 from Dmitry G. Dyachenko --- r266001 PASS
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 --- Comment #2 from Dmitry G. Dyachenko --- gcc invocation can be simplified : -fno-semantic-interposition can be removed. (was needed while creducing original testcase) So, $ g++ -fpreprocessed -O2 -flto -fPIC -DPIC -c -o a.o a.ii $ g++ -fpreprocessed -O2 -flto -fPIC -DPIC -c -o b.o b.ii $ g++ -flto -fPIC -DPIC -shared a.o b.o
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 Richard Biener changed: What|Removed |Added Target Milestone|--- |9.0
[Bug lto/87089] [9 regression] tree check: expected class 'type', have 'declaration' (namespace_decl) in type_with_linkage_p, at ipa-utils.h
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87089 Martin Liška changed: What|Removed |Added Status|UNCONFIRMED |ASSIGNED Last reconfirmed||2018-08-24 Assignee|unassigned at gcc dot gnu.org |hubicka at gcc dot gnu.org Ever confirmed|0 |1 --- Comment #1 from Martin Liška --- Confirmed, started with r263697.
