PR 84136 reports an ICE within sccvn_dom_walker when handling a
C/C++ source file that overuses the labels-as-values extension.
The code in question stores a jump label into a global, and then
jumps to it from another function, which ICEs after inlining:
void* a;
void foo() {
if ((a = &&l))
return;
l:;
}
int main() {
foo();
goto *a;
return 0;
}
This appears to be far beyond what we claim to support in this
extension - but we shouldn't ICE.
What's happening is that, after inlining, we have usage of a *copy*
of the label, which optimizes away the if-return logic, turning it
into an infinite loop.
On entry to the sccvn_dom_walker we have this gimple:
main ()
{
void * a.0_1;
<bb 2> [count: 0]:
a = &l;
<bb 3> [count: 0]:
l:
a.0_1 = a;
goto a.0_1;
}
and:
edge taken = find_taken_edge (bb, vn_valueize (val));
reasonably valueizes the:
goto a.0_1;
after the:
a = &l;
a.0_1 = a;
as if it were:
goto *&l;
find_taken_edge_computed_goto then has:
2380 dest = label_to_block (val);
2381 if (dest)
2382 {
2383 e = find_edge (bb, dest);
2384 gcc_assert (e != NULL);
2385 }
which locates dest as a self-jump from block 3 back to itself.
However, the find_edge call returns NULL - it has a predecessor edge
from block 2, but no successor edges.
Hence the assertion fails and we ICE.
A successor edge from the computed goto could have been created by
make_edges if the label stmt had been in the function, but make_edges
only looks in the current function when handling computed gotos, and
the label only appeared after inlining.
The following patch removes the assertion, fixing the ICE.
gcc/testsuite/ChangeLog:
PR tree-optimization/84136
* gcc.c-torture/compile/pr84136.c: New test.
gcc/ChangeLog:
PR tree-optimization/84136
* tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
that the result of find_edge is non-NULL.
From-SVN: r257509
+2018-02-08 David Malcolm <dmalcolm@redhat.com>
+
+ PR tree-optimization/84136
+ * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
+ that the result of find_edge is non-NULL.
+
2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
PR target/83008
+2018-02-08 David Malcolm <dmalcolm@redhat.com>
+
+ PR tree-optimization/84136
+ * gcc.c-torture/compile/pr84136.c: New test.
+
2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
PR target/83008
--- /dev/null
+void* a;
+
+void foo() {
+ if ((a = &&l))
+ return;
+
+ l:;
+}
+
+int main() {
+ foo();
+ goto *a;
+
+ return 0;
+}
dest = label_to_block (val);
if (dest)
- {
- e = find_edge (bb, dest);
- gcc_assert (e != NULL);
- }
+ e = find_edge (bb, dest);
+
+ /* It's possible for find_edge to return NULL here on invalid code
+ that abuses the labels-as-values extension (e.g. code that attempts to
+ jump *between* functions via stored labels-as-values; PR 84136).
+ If so, then we simply return that NULL for the edge.
+ We don't currently have a way of detecting such invalid code, so we
+ can't assert that it was the case when a NULL edge occurs here. */
return e;
}