[Bug ipa/92109] [10 Regression] ICE in modify_call_stmt, at ipa-param-manipulation.c:1586
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92109 Martin Jambor changed: What|Removed |Added Status|ASSIGNED|RESOLVED Resolution|--- |FIXED --- Comment #9 from Martin Jambor --- Fixed.
[Bug ipa/92109] [10 Regression] ICE in modify_call_stmt, at ipa-param-manipulation.c:1586
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92109 --- Comment #8 from Martin Jambor --- Author: jamborm Date: Mon Nov 25 10:13:08 2019 New Revision: 278670 URL: https://gcc.gnu.org/viewcvs?rev=278670=gcc=rev Log: ipa: Prevent materialization of clones with removed bodies (PR 92109) 2019-11-25 Martin Jambor PR ipa/92109 * cgraph.h (cgraph_node::remove_from_clone_tree): Declare. * cgraphclones.c (cgraph_node::remove_from_clone_tree): New method. (cgraph_materialize_clone): Move removel from clone tree to the the new method and use it instead. * ipa.c (symbol_table::remove_unreachable_nodes): When removing bodies of clones, also remove it from the clone tree. Modified: trunk/gcc/ChangeLog trunk/gcc/cgraph.h trunk/gcc/cgraphclones.c trunk/gcc/ipa.c
[Bug ipa/92109] [10 Regression] ICE in modify_call_stmt, at ipa-param-manipulation.c:1586
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92109 --- Comment #7 from Martin Jambor --- After talking to Honza, I have proposed a different patch on the mailing list: https://gcc.gnu.org/ml/gcc-patches/2019-11/msg02205.html
[Bug ipa/92109] [10 Regression] ICE in modify_call_stmt, at ipa-param-manipulation.c:1586
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92109
--- Comment #6 from Martin Jambor ---
So this helps:
diff --git a/gcc/cgraphclones.c b/gcc/cgraphclones.c
index bfcebb20495..359ea53d8a6 100644
--- a/gcc/cgraphclones.c
+++ b/gcc/cgraphclones.c
@@ -1079,6 +1079,7 @@ symbol_table::materialize_all_clones (void)
FOR_EACH_FUNCTION (node)
{
if (node->clone_of && node->decl != node->clone_of->decl
+ && !node->in_other_partition
&& !gimple_has_body_p (node->decl))
{
if (!node->clone_of->clone_of)
but I a doubt it is a safe thing to do (inline clones of a clone
probably can happen to be in a different ltrans than the offline clone
and then this will break) but it IMHO shows that the materialization
of this clone in this ltrans in this particular case is useless.
Hm, perhaps the following should work?
diff --git a/gcc/cgraphclones.c b/gcc/cgraphclones.c
index bfcebb20495..cc689d3d386 100644
--- a/gcc/cgraphclones.c
+++ b/gcc/cgraphclones.c
@@ -1079,6 +1079,7 @@ symbol_table::materialize_all_clones (void)
FOR_EACH_FUNCTION (node)
{
if (node->clone_of && node->decl != node->clone_of->decl
+ && (!node->body_removed || !node->in_other_partition)
&& !gimple_has_body_p (node->decl))
{
if (!node->clone_of->clone_of)
I'll see what breaks when I change the flag body_removed to mean what
it says (i.e. node->release_body really was called) first though.
[Bug ipa/92109] [10 Regression] ICE in modify_call_stmt, at ipa-param-manipulation.c:1586
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92109 Martin Jambor changed: What|Removed |Added CC||hubicka at gcc dot gnu.org --- Comment #5 from Martin Jambor --- It's because we are materializing: #7 0x010fbcb6 in tree_function_versioning ( old_decl=, new_decl=, tree_map=0x0, param_adjustments=0x7749fe20, update_clones=true, blocks_to_copy=0x0, new_entry=0x0) at /home/mjambor/gcc/icln/src/gcc/tree-inline.c:6279 6279 copy_body (, ENTRY_BLOCK_PTR_FOR_FN (cfun), EXIT_BLOCK_PTR_FOR_FN (cfun), (gdb) p new_version_node->debug() _ZN3KJS10throwErrorEPNS_9ExecStateENS_9ErrorTypeERK7QString.constprop.0.isra.0/1877 (throwError.constprop.isra) @0x775179a0 Type: function Body removed by symtab_remove_unreachable_nodes ^ this. And the constprop clone is gone... and we act as if that was the body from which we create a version... and that does not work.
[Bug ipa/92109] [10 Regression] ICE in modify_call_stmt, at ipa-param-manipulation.c:1586
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92109 Martin Liška changed: What|Removed |Added Keywords|needs-reduction | Assignee|marxin at gcc dot gnu.org |jamborm at gcc dot gnu.org --- Comment #4 from Martin Liška --- I can't reduce it more, so it's hopefully a reasonable small test-case: $ g++ 1.ii 2.ii 3.ii 4.ii -O3 -flto=auto -shared -fPIC ... 1.ii:42:572: note: code may be misoptimized unless ‘-fno-strict-aliasing’ is used during IPA pass: materialize-all-clones In function ‘throwError.constprop.isra’: lto1: internal compiler error: in modify_call_stmt, at ipa-param-manipulation.c:1586 0x59f0b1 ipa_param_body_adjustments::modify_call_stmt(gcall**) ../../gcc/ipa-param-manipulation.c:1586 0xa646b5 remap_gimple_stmt ../../gcc/tree-inline.c:1921 0xa6588f copy_bb ../../gcc/tree-inline.c:1996 0xa66de2 copy_cfg_body ../../gcc/tree-inline.c:2983 0xa66de2 copy_body ../../gcc/tree-inline.c:3231 0xa6ba0c tree_function_versioning(tree_node*, tree_node*, vec*, ipa_param_adjustments*, bool, bitmap_head*, basic_block_def*) ../../gcc/tree-inline.c:6274 0x67a20a cgraph_materialize_clone ../../gcc/cgraphclones.c:1020 0x67a20a symbol_table::materialize_all_clones() ../../gcc/cgraphclones.c:1107 0x85af5f execute ../../gcc/ipa.c:1396
[Bug ipa/92109] [10 Regression] ICE in modify_call_stmt, at ipa-param-manipulation.c:1586
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92109 --- Comment #3 from Martin Liška --- Created attachment 47088 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=47088=edit Reduced test-case
[Bug ipa/92109] [10 Regression] ICE in modify_call_stmt, at ipa-param-manipulation.c:1586
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92109 Martin Liška changed: What|Removed |Added CC||jamborm at gcc dot gnu.org Known to work||9.2.0 Known to fail||10.0 --- Comment #2 from Martin Liška --- Ok, now I have just isolated 4 pre-processed source files and I'm reducing these ..
[Bug ipa/92109] [10 Regression] ICE in modify_call_stmt, at ipa-param-manipulation.c:1586
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92109 Martin Liška changed: What|Removed |Added Keywords||needs-reduction Status|UNCONFIRMED |ASSIGNED Last reconfirmed||2019-10-22 Assignee|unassigned at gcc dot gnu.org |marxin at gcc dot gnu.org Ever confirmed|0 |1 --- Comment #1 from Martin Liška --- Let me work on test-case isolation.
[Bug ipa/92109] [10 Regression] ICE in modify_call_stmt, at ipa-param-manipulation.c:1586
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92109 Richard Biener changed: What|Removed |Added Target Milestone|--- |10.0 Summary|ICE in modify_call_stmt, at |[10 Regression] ICE in |ipa-param-manipulation.c:15 |modify_call_stmt, at |86 |ipa-param-manipulation.c:15 ||86
