With revision 163380 this test passed. From revision 163383 and on, this test has failed as follows: Running /tmp/regressn1-383/gcc/gcc/testsuite/gcc.c-torture/execute/execute.exp ... FAIL: gcc.c-torture/execute/990326-1.c execution, -O2 FAIL: gcc.c-torture/execute/990326-1.c execution, -O3 -fomit-frame-pointer FAIL: gcc.c-torture/execute/990326-1.c execution, -O3 -g FAIL: gcc.c-torture/execute/990326-1.c execution, -Os FAIL: gcc.c-torture/execute/990326-1.c execution, -O2 -flto FAIL: gcc.c-torture/execute/990326-1.c execution, -O2 -fwhopr With the message in the logfile showing that abort was called. RTL dumps show that the misoptimizing is done in the combine pass. A simplified version of gcc.c-torture/execute/990326-1.c exposing the miscompilation will be attached. Author of the suspect patch CC:ed.
Created attachment 21707 [details] Simplified gcc.c-torture/execute/990326-1.c See the combine dump at e.g. -O2. I haven't disected the miscompilation, but at a glance it looks like the subregness is not handled properly.
Fixed by r163917, apparently.