https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126222
Bug ID: 126222
Summary: ranger: threading path containing pointer inequality
to resolved
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: aldyh at gcc dot gnu.org
CC: amacleod at redhat dot com
Depends on: 125986
Blocks: 126010
Target Milestone: ---
[Reproduce with -O2 -fdump-tree-dom-all. I am attaching the .ii file.]
For diagnostics/sarif-sink.ii, distilled from a stage1 bootstrap, the DOM
tables are resolving a conditional on a path which ranger cannot. I am
attaching a couple patches to reproduce this (the same as I did for PR126212).
With these we ICE while trying to unsuccesfully fold (DOM resolves it to true):
if (search$_M_node_151 != &MEM[(struct _Rb_tree *)&arr +
40B]._M_impl.D.319560._M_header)
The threading candidate path is BB18->BB20.
BB13:
if (__y_177 == &MEM[(struct _Rb_tree *)&arr +
40B]._M_impl.D.319560._M_header)
blah
else
goto bb15
...
... (BB15 dominates BB18)
...
BB18:
;; __y_177 != &MEM[....] is known on entry
if (_182 < 0)
blah
else
goto bb20
BB20:
cstore_554 = PHI <&MEM[(struct _Rb_tree *)&arr +
40B]._M_impl.D.319560._M_header(5), __y_177(18)>
SR.5218_179 = cstore_554
SR.5216_178 = SR.5218_179
search$_M_node_151 = SR.5216_178
;;
;; After the above copies, this amounts to __y_177 != &MEM[...]
;; VVVVVVV
if (search$_M_node_151 != &MEM[(struct _Rb_tree *)&arr +
40B]._M_impl.D.319560._M_header)
So clearly we should know how to fold the final conditional in this path.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125986
[Bug 125986] ranger: incoming relations to a threading path are ignored
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126010
[Bug 126010] [meta-bug] Remove DOM
