| Message ID | 87r0kiu7f3.fsf@euler.schwinge.homeip.net |
|---|---|
| State | New |
| Headers | show |
| Series | Fix 'gcc.dg/tree-ssa/return-value-range-1.c' for 'char' defaulting to 'unsigned' (was: Propagate value ranges of return values) | expand |
Hi! On Tue, 21 Nov 2023 at 22:24, Thomas Schwinge <thomas@codesourcery.com> wrote: > > Hi! > > On 2023-11-19T16:05:42+0100, Jan Hubicka <hubicka@ucw.cz> wrote: > > --- /dev/null > > +++ b/gcc/testsuite/gcc.dg/tree-ssa/return-value-range-1.c > > Pushed to master branch commit a0240662b22312ffb3e3fefb85f258ab0e7010f4 > "Fix 'gcc.dg/tree-ssa/return-value-range-1.c' for 'char' defaulting to > 'unsigned'", see attached. On powerpc64le-linux-gnu ('char' defaulting > to 'unsigned') I still saw: > > /tmp/ccd1xwD7.o: In function `test': > return-value-range-1.c:(.text+0x50): undefined reference to `link_error' > We do see the same error in our CI (Thomas, normally you have received a notification because your patch turned ERROR in FAIL) Thomas, you said in another email that adding -O2 avoids the linker error with missing link_error(), but I don't see how that would be possible? (and hence I expect the error you quoted above to happen) So should we use dg-compile instead of dg-link? Not sure what the original intention was? Thanks, Christophe > > Grüße > Thomas > > > > @@ -0,0 +1,22 @@ > > +/* { dg-do ling } */ > > +/* { dg-options "-O1 -dump-tree-evrp-details" } */ > > +__attribute__ ((__noinline__)) > > +int a(char c) > > +{ > > + return c; > > +} > > +void link_error (); > > + > > +void > > +test(int d) > > +{ > > + if (a(d) > 200) > > + link_error (); > > +} > > +int > > +main(int argc, char **argv) > > +{ > > + test(argc); > > + return 0; > > +} > > +/* { dg-final { scan-tree-dump-times "Recording return range" 2 "evrp"} } */ > > > ----------------- > Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht München, HRB 106955
Hi! On 2023-11-22T11:51:02+0100, Christophe Lyon <christophe.lyon@linaro.org> wrote: > On Tue, 21 Nov 2023 at 22:24, Thomas Schwinge <thomas@codesourcery.com> wrote: >> On 2023-11-19T16:05:42+0100, Jan Hubicka <hubicka@ucw.cz> wrote: >> > --- /dev/null >> > +++ b/gcc/testsuite/gcc.dg/tree-ssa/return-value-range-1.c >> >> Pushed to master branch commit a0240662b22312ffb3e3fefb85f258ab0e7010f4 >> "Fix 'gcc.dg/tree-ssa/return-value-range-1.c' for 'char' defaulting to >> 'unsigned'", see attached. On powerpc64le-linux-gnu ('char' defaulting >> to 'unsigned') I still saw: >> >> /tmp/ccd1xwD7.o: In function `test': >> return-value-range-1.c:(.text+0x50): undefined reference to `link_error' >> > We do see the same error in our CI (Thomas, normally you have received > a notification because your patch turned ERROR in FAIL) Yes, I have; and I even tried to log in there, to point to my commit mentioned above, which is meant to address this issue -- please let me know if you're still seeing the FAIL after that commit. > Thomas, you said in another email that adding -O2 avoids the linker > error with missing link_error(), but I don't see how that would be > possible? That's the gist of Honza's "Propagate value ranges of return values" optimization, per my understanding: from 'int a(signed char c)' doing 'return c;' figure out that 'a(d) > 200)' is always false (due to '-128 <= c <= 127)'. > (and hence I expect the error you quoted above to happen) > > So should we use dg-compile instead of dg-link? Not sure what the > original intention was? No, the idea really is to prove that the 'link_error ()' call is unreachable. Grüße Thomas >> > @@ -0,0 +1,22 @@ >> > +/* { dg-do ling } */ >> > +/* { dg-options "-O1 -dump-tree-evrp-details" } */ >> > +__attribute__ ((__noinline__)) >> > +int a(char c) >> > +{ >> > + return c; >> > +} >> > +void link_error (); >> > + >> > +void >> > +test(int d) >> > +{ >> > + if (a(d) > 200) >> > + link_error (); >> > +} >> > +int >> > +main(int argc, char **argv) >> > +{ >> > + test(argc); >> > + return 0; >> > +} >> > +/* { dg-final { scan-tree-dump-times "Recording return range" 2 "evrp"} } */ ----------------- Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht München, HRB 106955
From a0240662b22312ffb3e3fefb85f258ab0e7010f4 Mon Sep 17 00:00:00 2001 From: Thomas Schwinge <thomas@codesourcery.com> Date: Tue, 21 Nov 2023 22:07:13 +0100 Subject: [PATCH] Fix 'gcc.dg/tree-ssa/return-value-range-1.c' for 'char' defaulting to 'unsigned' ... added in recent commit 53ba8d669550d3a1f809048428b97ca607f95cf5 "inter-procedural value range propagation", fixed in commit 878a860cae78146d98d7a21612f0bcec0930a9c2 "Fix 'gcc.dg/tree-ssa/return-value-range-1.c'". gcc/testsuite/ * gcc.dg/tree-ssa/return-value-range-1.c: Fix. --- gcc/testsuite/gcc.dg/tree-ssa/return-value-range-1.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/gcc/testsuite/gcc.dg/tree-ssa/return-value-range-1.c b/gcc/testsuite/gcc.dg/tree-ssa/return-value-range-1.c index 74f1a5080bb..97294482c05 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/return-value-range-1.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/return-value-range-1.c @@ -1,7 +1,7 @@ /* { dg-do link } */ /* { dg-options "-O2 -fdump-tree-evrp-details" } */ __attribute__ ((__noinline__)) -int a(char c) +int a(signed char c) { return c; } -- 2.34.1