[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763 --- Comment #13 from Jan Hubicka --- Author: hubicka Date: Tue Jan 9 08:22:28 2018 New Revision: 256369 URL: https://gcc.gnu.org/viewcvs?rev=256369=gcc=rev Log: PR ipa/80763 * ipa-comdats.c (set_comdat_group): Only set comdat group of real symbols; not inline clones. Modified: trunk/gcc/ChangeLog trunk/gcc/ipa-comdats.c
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763
Jan Hubicka changed:
What|Removed |Added
Assignee|marxin at gcc dot gnu.org |hubicka at gcc dot
gnu.org
--- Comment #12 from Jan Hubicka ---
Sorry, attached incorrect path.
Index: ipa-comdats.c
===
--- ipa-comdats.c (revision 256232)
+++ ipa-comdats.c (working copy)
@@ -211,8 +211,11 @@ set_comdat_group (symtab_node *symbol,
symtab_node *head = (symtab_node *)head_p;
gcc_assert (!symbol->get_comdat_group ());
- symbol->set_comdat_group (head->get_comdat_group ());
- symbol->add_to_same_comdat_group (head);
+ if (symbol->real_symbol_p ())
+{
+ symbol->set_comdat_group (head->get_comdat_group ());
+ symbol->add_to_same_comdat_group (head);
+}
return false;
}
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763
Jan Hubicka changed:
What|Removed |Added
Status|NEW |ASSIGNED
--- Comment #11 from Jan Hubicka ---
The reduced testcase does not reproduce anymore so I assumed this to be fixed,
sorry for that. The problem is that comdats pass privatizes inlined thunk.
This is bid odd side case of for_symbol_thunk_and_aliases - honestly i am not
sure if it should consider inlined thunk to be thunk anymore, but given the
name it probably should.
Honza
Index: ipa-comdats.c
===
--- ipa-comdats.c (revision 256232)
+++ ipa-comdats.c (working copy)
@@ -341,7 +341,7 @@ ipa_comdats (void)
if (!symbol->get_comdat_group ()
&& !symbol->alias
&& (!(fun = dyn_cast (symbol))
- || !fun->thunk.thunk_p)
+ || (!fun->thunk.thunk_p && !fun->global.inlined_to))
&& symbol->real_symbol_p ())
{
tree *val = map.get (symbol);
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763 David Binderman changed: What|Removed |Added CC||rguenth at gcc dot gnu.org --- Comment #10 from David Binderman --- (In reply to David Binderman from comment #9) > Broken for over four months now. Still broken. Regression from 8.0 Since hubicka hasn't fixed it in six months, perhaps it is time to allow someone else to have a go ? I'd be happy to help with any testing of any speculative patch.
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763 --- Comment #9 from David Binderman --- (In reply to David Binderman from comment #8) > (In reply to David Binderman from comment #7) > > Still seems to be broken, over a month later. > > Still broken, a couple of months even later ... Broken for over four months now.
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763 --- Comment #8 from David Binderman --- (In reply to David Binderman from comment #7) > Still seems to be broken, over a month later. Still broken, a couple of months even later ...
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763 Richard Biener changed: What|Removed |Added Target Milestone|7.2 |7.3
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763 Richard Biener changed: What|Removed |Added Target Milestone|7.2 |7.3 --- Comment #11 from Richard Biener --- GCC 7.2 is being released, adjusting target milestone. --- Comment #12 from Richard Biener --- GCC 7.2 is being released, adjusting target milestone.
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763 Richard Biener changed: What|Removed |Added Target Milestone|7.2 |7.3 --- Comment #11 from Richard Biener --- GCC 7.2 is being released, adjusting target milestone. --- Comment #12 from Richard Biener --- GCC 7.2 is being released, adjusting target milestone.
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763 --- Comment #10 from Martin Liška --- Hm, the function clone_inlined_nodes contains logic that does: │246 e->callee->remove_from_same_comdat_group (); but only when duplicate == false. Will be more complex to handle, please take a look Honza.
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763
Martin Liška changed:
What|Removed |Added
CC||marxin at gcc dot gnu.org
--- Comment #8 from Martin Liška ---
Created attachment 41987
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=41987=edit
Patch candidate
I wrongly pasted 2 patches to a different PR.
Adding patch candidate that changes cgraph_node::verify_node that verifies
global.inlined_to && same_comdat_group. It believe proper fix is to verify that
node should live in a same comdat group (if belongs to any).
Honza's reply:
OK, so now the tests passes when inline clone is either in no comdat group list
(even if function inlined to is in one) or if the comdat groups are the same.
Does it fail when we test that comdats groups are always the same? It seems
iffy to allow silently both. I guess both variants are fine, but lets pick one
of them. Perhaps we want only to remove node from comdat group list while
inlining it.
My reply:
I'm going to test it whether more strict test will survive.
Honza
--- Comment #9 from Martin Liška ---
So having:
if (global.inlined_to
&& get_comdat_group () != global.inlined_to->get_comdat_group ())
{
error ("inline clone in a different comdat group list");
error_found = true;
}
fails in situation like this:
namespace {
class a
{
public:
a ();
};
class b : a
{
};
}
namespace c {
class e : b
{
};
d () { e (); }
}
ice.i:12:7: error: inline clone in a different comdat group list
class e : b
^
_ZN12_GLOBAL__N_11bC2Ev/1 ({anonymous}::b::b()) @0x2b14a0ce4170
Type: function definition analyzed
Visibility: prevailing_def_ironly artificial
References:
Referring:
Function {anonymous}::b::b()/1 is inline copy in c::e::e()/4
Availability: local
First run: 0
Function flags: body local
Called by: c::e::e()/4 (inlined) (1.00 per call) (can throw external)
Calls: {anonymous}::a::a()/7 (1.00 per call) (can throw external)
during GIMPLE pass: einline
ice.i:12:7: internal compiler error: verify_cgraph_node failed
0x979eb3 cgraph_node::verify_node()
../../gcc/cgraph.c:3537
0x96e15c symtab_node::verify()
../../gcc/symtab.c:1204
0xe241b9 expand_call_inline
../../gcc/tree-inline.c:4581
0xe27184 gimple_expand_calls_inline
../../gcc/tree-inline.c:4954
0xe27184 optimize_inline_calls(tree_node*)
../../gcc/tree-inline.c:5094
0x1556a39 early_inliner(function*)
../../gcc/ipa-inline.c:2727
So do you prefer changing the comdat group when inlining or original version of
verifier is preferred?
Honza:
I would say that we have loops walking all symbols in given comdat group that
are not
really expecting to trip over inline clones, so it may be easier to
consistently take
inline clones out of the comdat group rather than consistently to put them in.
Honza
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763 Martin Liška changed: What|Removed |Added CC||marxin at gcc dot gnu.org --- Comment #8 from Martin Liška --- Created attachment 41987 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=41987=edit Patch candidate I wrongly pasted 2 patches to a different PR. Adding patch candidate that changes cgraph_node::verify_node that verifies global.inlined_to && same_comdat_group. It believe proper fix is to verify that node should live in a same comdat group (if belongs to any). Honza's reply: OK, so now the tests passes when inline clone is either in no comdat group list (even if function inlined to is in one) or if the comdat groups are the same. Does it fail when we test that comdats groups are always the same? It seems iffy to allow silently both. I guess both variants are fine, but lets pick one of them. Perhaps we want only to remove node from comdat group list while inlining it. My reply: I'm going to test it whether more strict test will survive. Honza
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763 --- Comment #7 from David Binderman --- Still seems to be broken, over a month later.
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763
--- Comment #6 from H.J. Lu ---
This is introduced by r236012 with a different ICE:
pr80763.cc:27:14: internal compiler error: in inline_small_functions, at
ipa-inline.c:1881
void n::o() {}
^
Please submit a full bug report,
with preprocessed source if appropriate.
See for instructions.
r236271 changed ICE to
pr80763.cc:27:14: error: inline clone in same comdat group list
void n::o() {}
^
_ZThn8_N1n1fE.artificial_thunk.2/42 () @0x7f2be1015e60
Type: function definition analyzed
Visibility: comdat_group:_ZN1iI1lE1dEjj artificial
Same comdat group as: _ZN1iI1lE1dEjj/25
References:
Referring:
Function /42 is inline copy in unsigned int i::d(unsigned int,
unsigned int) [with e = l]/25
Availability: local
First run: 0
Function flags: local nonfreeing_fn
Thunk fixed offset -8 virtual value 0 has virtual offset 0)
Called by: _ZN1iI1lE1dEjj.part.0/35 (speculative) (inlined) (0.31 per call)
(can throw external)
Calls: _ZN1n1fE.constprop.1/41 (call_stmt_cannot_inline_p) (0.31 per
call)
pr80763.cc:27:14: internal compiler error: verify_cgraph_node failed
Please submit a full bug report,
with preprocessed source if appropriate.
See for instructions.
gnu-ivb-1:pts/1[63]> ./gcc-test-run 236267
Testing /export/gnu/import/git/gcc-regression/master/236267/usr/bin/gcc:
pr80763.cc:27:14: internal compiler error: in inline_small_functions, at
ipa-inline.c:1881
void n::o() {}
^
Please submit a full bug report,
with preprocessed source if appropriate.
See for instructions.
[Bug c++/80763] [7/8 Regression] -O3 causes error: inline clone in same comdat group list
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80763 Richard Biener changed: What|Removed |Added Priority|P3 |P2 Status|UNCONFIRMED |NEW Known to work||6.3.0 Keywords||ice-checking Last reconfirmed||2017-05-16 CC||hubicka at gcc dot gnu.org Ever confirmed|0 |1 Summary|-O3 causes error: inline|[7/8 Regression] -O3 causes |clone in same comdat group |error: inline clone in same |list|comdat group list Target Milestone|--- |7.2 Known to fail||7.1.0 --- Comment #5 from Richard Biener --- Confirmed. Fails on the branch with -fchecking.
