| Message ID | 20180724095027.GC17988@tucnak |
|---|---|
| State | New |
| Headers | show |
| Series | Fix up pr19476-{1,5}.C (PR testsuite/86649) | expand |
On Tue, 24 Jul 2018, Jakub Jelinek wrote: > Hi! > > When looking at PR86569 testresults, I must have missed these two tests > (but looking at test_summary outputs, I see it now). > When we no longer fold this during cp_fold (to avoid code generation > changes between -Wnonnull-compare and -Wno-nonnull-compare), it isn't > folded from the first pass; with -O2 it is folded during evrp and with > -O1 during dom2. > > Note, the test would fail before with -Wnonnull-compare, e.g. on 8 > branch (which doesn't have the PR86569 changes), I see: > make check-c++-all RUNTESTFLAGS='--target_board=unix\{,-Wnonnull-compare\} dg.exp=pr19476*' > === g++ Summary for unix === > > # of expected passes 72 > Running target unix/-Wnonnull-compare > Using /usr/share/dejagnu/baseboards/unix.exp as board description file for target. > Using /usr/share/dejagnu/config/unix.exp as generic interface file for target. > Using /usr/src/gcc-8/gcc/testsuite/config/default.exp as tool-and-target-specific interface file. > Running /usr/src/gcc-8/gcc/testsuite/g++.dg/dg.exp ... > FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++98 scan-tree-dump ccp1 "return 42" > FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++11 scan-tree-dump ccp1 "return 42" > FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++14 scan-tree-dump ccp1 "return 42" > FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++17 scan-tree-dump ccp1 "return 42" > FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++2a scan-tree-dump ccp1 "return 42" > FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++17 -fconcepts scan-tree-dump ccp1 "return 42" > FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++98 scan-tree-dump ccp1 "return 42" > FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++11 scan-tree-dump ccp1 "return 42" > FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++14 scan-tree-dump ccp1 "return 42" > FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++17 scan-tree-dump ccp1 "return 42" > FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++2a scan-tree-dump ccp1 "return 42" > FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++17 -fconcepts scan-tree-dump ccp1 "return 42" > > === g++ Summary for unix/-Wnonnull-compare === > > # of expected passes 60 > # of unexpected failures 12 > > Especially for -O2 that people use most, folding it at evrp time seems to be > early enough for me. > Fixed by testing this only in dom2, tested on x86_64-linux, ok for trunk? OK - can you add a variant with -O2 that tests it at EVRP time then? Thanks, Richard. > 2018-07-24 Jakub Jelinek <jakub@redhat.com> > > PR testsuite/86649 > * g++.dg/tree-ssa-/pr19476-1.C: Check dom2 dump instead of ccp1. > * g++.dg/tree-ssa-/pr19476-5.C: Likewise. > > --- gcc/testsuite/g++.dg/tree-ssa/pr19476-1.C.jj 2015-05-29 15:04:33.037803445 +0200 > +++ gcc/testsuite/g++.dg/tree-ssa/pr19476-1.C 2018-07-24 11:39:10.108897097 +0200 > @@ -1,5 +1,5 @@ > /* { dg-do compile } */ > -/* { dg-options "-O -fdump-tree-ccp1 -fdelete-null-pointer-checks" } */ > +/* { dg-options "-O -fdump-tree-dom2 -fdelete-null-pointer-checks" } */ > /* { dg-skip-if "" keeps_null_pointer_checks } */ > > // See pr19476-5.C for a version without including <new>. > @@ -12,5 +12,5 @@ int g(){ > return 42 + (0 == new int[50]); > } > > -/* { dg-final { scan-tree-dump "return 42" "ccp1" } } */ > -/* { dg-final { scan-tree-dump-not "return 33" "ccp1" } } */ > +/* { dg-final { scan-tree-dump "return 42" "dom2" } } */ > +/* { dg-final { scan-tree-dump-not "return 33" "dom2" } } */ > --- gcc/testsuite/g++.dg/tree-ssa/pr19476-5.C.jj 2015-05-29 15:04:33.038803430 +0200 > +++ gcc/testsuite/g++.dg/tree-ssa/pr19476-5.C 2018-07-24 11:39:26.190913802 +0200 > @@ -1,5 +1,5 @@ > /* { dg-do compile } */ > -/* { dg-options "-O -fdump-tree-ccp1 -fdelete-null-pointer-checks" } */ > +/* { dg-options "-O -fdump-tree-dom2 -fdelete-null-pointer-checks" } */ > /* { dg-skip-if "" keeps_null_pointer_checks } */ > > // See pr19476-1.C for a version that includes <new>. > @@ -8,4 +8,4 @@ int g(){ > return 42 + (0 == new int[50]); > } > > -/* { dg-final { scan-tree-dump "return 42" "ccp1" } } */ > +/* { dg-final { scan-tree-dump "return 42" "dom2" } } */ > > Jakub > >
On Tue, Jul 24, 2018 at 12:08:35PM +0200, Richard Biener wrote:
> OK - can you add a variant with -O2 that tests it at EVRP time then?
Here is what I've committed to trunk then:
2018-07-24 Jakub Jelinek <jakub@redhat.com>
PR testsuite/86649
* g++.dg/tree-ssa-/pr19476-1.C: Check dom2 dump instead of ccp1.
* g++.dg/tree-ssa-/pr19476-5.C: Likewise.
* g++.dg/tree-ssa-/pr19476-6.C: New test.
* g++.dg/tree-ssa-/pr19476-7.C: New test.
--- gcc/testsuite/g++.dg/tree-ssa/pr19476-1.C.jj 2015-05-29 15:04:33.037803445 +0200
+++ gcc/testsuite/g++.dg/tree-ssa/pr19476-1.C 2018-07-24 11:39:10.108897097 +0200
@@ -1,5 +1,5 @@
/* { dg-do compile } */
-/* { dg-options "-O -fdump-tree-ccp1 -fdelete-null-pointer-checks" } */
+/* { dg-options "-O -fdump-tree-dom2 -fdelete-null-pointer-checks" } */
/* { dg-skip-if "" keeps_null_pointer_checks } */
// See pr19476-5.C for a version without including <new>.
@@ -12,5 +12,5 @@ int g(){
return 42 + (0 == new int[50]);
}
-/* { dg-final { scan-tree-dump "return 42" "ccp1" } } */
-/* { dg-final { scan-tree-dump-not "return 33" "ccp1" } } */
+/* { dg-final { scan-tree-dump "return 42" "dom2" } } */
+/* { dg-final { scan-tree-dump-not "return 33" "dom2" } } */
--- gcc/testsuite/g++.dg/tree-ssa/pr19476-5.C.jj 2015-05-29 15:04:33.038803430 +0200
+++ gcc/testsuite/g++.dg/tree-ssa/pr19476-5.C 2018-07-24 11:39:26.190913802 +0200
@@ -1,5 +1,5 @@
/* { dg-do compile } */
-/* { dg-options "-O -fdump-tree-ccp1 -fdelete-null-pointer-checks" } */
+/* { dg-options "-O -fdump-tree-dom2 -fdelete-null-pointer-checks" } */
/* { dg-skip-if "" keeps_null_pointer_checks } */
// See pr19476-1.C for a version that includes <new>.
@@ -8,4 +8,4 @@ int g(){
return 42 + (0 == new int[50]);
}
-/* { dg-final { scan-tree-dump "return 42" "ccp1" } } */
+/* { dg-final { scan-tree-dump "return 42" "dom2" } } */
--- gcc/testsuite/g++.dg/tree-ssa/pr19476-6.C.jj 2018-07-24 12:09:30.321890628 +0200
+++ gcc/testsuite/g++.dg/tree-ssa/pr19476-6.C 2018-07-24 12:10:49.812987922 +0200
@@ -0,0 +1,16 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-evrp -fdelete-null-pointer-checks" } */
+/* { dg-skip-if "" keeps_null_pointer_checks } */
+
+// See pr19476-7.C for a version without including <new>.
+#include <new>
+
+int f(){
+ return 33 + (0 == new(std::nothrow) int);
+}
+int g(){
+ return 42 + (0 == new int[50]);
+}
+
+/* { dg-final { scan-tree-dump "return 42" "evrp" } } */
+/* { dg-final { scan-tree-dump-not "return 33" "evrp" } } */
--- gcc/testsuite/g++.dg/tree-ssa/pr19476-7.C.jj 2018-07-24 12:09:33.034893945 +0200
+++ gcc/testsuite/g++.dg/tree-ssa/pr19476-7.C 2018-07-24 12:11:03.657004866 +0200
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-evrp -fdelete-null-pointer-checks" } */
+/* { dg-skip-if "" keeps_null_pointer_checks } */
+
+// See pr19476-6.C for a version that includes <new>.
+
+int g(){
+ return 42 + (0 == new int[50]);
+}
+
+/* { dg-final { scan-tree-dump "return 42" "evrp" } } */
Jakub
--- gcc/testsuite/g++.dg/tree-ssa/pr19476-1.C.jj 2015-05-29 15:04:33.037803445 +0200 +++ gcc/testsuite/g++.dg/tree-ssa/pr19476-1.C 2018-07-24 11:39:10.108897097 +0200 @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-O -fdump-tree-ccp1 -fdelete-null-pointer-checks" } */ +/* { dg-options "-O -fdump-tree-dom2 -fdelete-null-pointer-checks" } */ /* { dg-skip-if "" keeps_null_pointer_checks } */ // See pr19476-5.C for a version without including <new>. @@ -12,5 +12,5 @@ int g(){ return 42 + (0 == new int[50]); } -/* { dg-final { scan-tree-dump "return 42" "ccp1" } } */ -/* { dg-final { scan-tree-dump-not "return 33" "ccp1" } } */ +/* { dg-final { scan-tree-dump "return 42" "dom2" } } */ +/* { dg-final { scan-tree-dump-not "return 33" "dom2" } } */ --- gcc/testsuite/g++.dg/tree-ssa/pr19476-5.C.jj 2015-05-29 15:04:33.038803430 +0200 +++ gcc/testsuite/g++.dg/tree-ssa/pr19476-5.C 2018-07-24 11:39:26.190913802 +0200 @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-O -fdump-tree-ccp1 -fdelete-null-pointer-checks" } */ +/* { dg-options "-O -fdump-tree-dom2 -fdelete-null-pointer-checks" } */ /* { dg-skip-if "" keeps_null_pointer_checks } */ // See pr19476-1.C for a version that includes <new>. @@ -8,4 +8,4 @@ int g(){ return 42 + (0 == new int[50]); } -/* { dg-final { scan-tree-dump "return 42" "ccp1" } } */ +/* { dg-final { scan-tree-dump "return 42" "dom2" } } */
Hi! When looking at PR86569 testresults, I must have missed these two tests (but looking at test_summary outputs, I see it now). When we no longer fold this during cp_fold (to avoid code generation changes between -Wnonnull-compare and -Wno-nonnull-compare), it isn't folded from the first pass; with -O2 it is folded during evrp and with -O1 during dom2. Note, the test would fail before with -Wnonnull-compare, e.g. on 8 branch (which doesn't have the PR86569 changes), I see: make check-c++-all RUNTESTFLAGS='--target_board=unix\{,-Wnonnull-compare\} dg.exp=pr19476*' === g++ Summary for unix === # of expected passes 72 Running target unix/-Wnonnull-compare Using /usr/share/dejagnu/baseboards/unix.exp as board description file for target. Using /usr/share/dejagnu/config/unix.exp as generic interface file for target. Using /usr/src/gcc-8/gcc/testsuite/config/default.exp as tool-and-target-specific interface file. Running /usr/src/gcc-8/gcc/testsuite/g++.dg/dg.exp ... FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++98 scan-tree-dump ccp1 "return 42" FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++11 scan-tree-dump ccp1 "return 42" FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++14 scan-tree-dump ccp1 "return 42" FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++17 scan-tree-dump ccp1 "return 42" FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++2a scan-tree-dump ccp1 "return 42" FAIL: g++.dg/tree-ssa/pr19476-1.C -std=gnu++17 -fconcepts scan-tree-dump ccp1 "return 42" FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++98 scan-tree-dump ccp1 "return 42" FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++11 scan-tree-dump ccp1 "return 42" FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++14 scan-tree-dump ccp1 "return 42" FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++17 scan-tree-dump ccp1 "return 42" FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++2a scan-tree-dump ccp1 "return 42" FAIL: g++.dg/tree-ssa/pr19476-5.C -std=gnu++17 -fconcepts scan-tree-dump ccp1 "return 42" === g++ Summary for unix/-Wnonnull-compare === # of expected passes 60 # of unexpected failures 12 Especially for -O2 that people use most, folding it at evrp time seems to be early enough for me. Fixed by testing this only in dom2, tested on x86_64-linux, ok for trunk? 2018-07-24 Jakub Jelinek <jakub@redhat.com> PR testsuite/86649 * g++.dg/tree-ssa-/pr19476-1.C: Check dom2 dump instead of ccp1. * g++.dg/tree-ssa-/pr19476-5.C: Likewise. Jakub