Created attachment 58487 [details] Reproducer Compiler in use: gnat-riscv64-elf-darwin-14.1.0-3 from https://github.com/alire-project/GNAT-FSF-builds/releases Compiling the attached reproducer with riscv64-elf-gcc -c --RTS=. bug.adb results in during RTL pass: expand +===========================GNAT BUG DETECTED==============================+ | 14.1.0 (riscv64-elf) in tree_to_uhwi, at tree.h:4951 | | Error detected around bug.adb:29:24 | | Compiling bug.adb | | Please submit a bug report; see https://gcc.gnu.org/bugs/ . | | Use a subject line meaningful to you and us to track the bug. | | Include the entire contents of this bug box in the report. | | Include the exact command that you entered. | | Also include sources listed below. | +==========================================================================+ Please include these source files with error report Note that list may not be accurate in some cases, so please double check that the problem can still be reproduced with the set of files listed. Consider also -gnatd.n switch (see debug.adb). system.ads bug.adb bug.ads s-stoele.ads s-stoele.adb ada.ads a-unccon.ads raised TYPES.UNRECOVERABLE_ERROR : comperr.adb:412 Note that * this error doesn’t happen with an arm-eabi cross compiler * it was present from GCC 12.2.0
It's a buglet in the RISC-V back-end: (gdb) bt #0 internal_error (gmsgid=0x584aa67 "in %s, at %s:%d") at /home/eric/cvs/gcc/gcc/diagnostic.cc:2255 #1 0x0000000004ad30a2 in fancy_abort ( file=0x4ea1be0 "/home/eric/cvs/gcc/gcc/tree.cc", line=6488, function=0x4ea3cc1 "tree_to_uhwi") at /home/eric/cvs/gcc/gcc/diagnostic.cc:2380 #2 0x0000000002469386 in tree_to_uhwi (t=0x7ffff705aca8) at /home/eric/cvs/gcc/gcc/tree.cc:6488 #3 0x0000000002555b2e in riscv_valid_lo_sum_p (sym_type=SYMBOL_ABSOLUTE, mode=E_BLKmode, x=0x7ffff717bb58) at /home/eric/cvs/gcc/gcc/config/riscv/riscv.cc:1706
The immediate fix is: diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index c17141d909a..5e34dc92210 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1702,7 +1702,9 @@ riscv_valid_lo_sum_p (enum riscv_symbol_type sym_type, machine_mode mode, align = (SYMBOL_REF_DECL (x) ? DECL_ALIGN (SYMBOL_REF_DECL (x)) : 1); - size = (SYMBOL_REF_DECL (x) && DECL_SIZE (SYMBOL_REF_DECL (x)) + size = (SYMBOL_REF_DECL (x) + && DECL_SIZE (SYMBOL_REF_DECL (x)) + && tree_fits_uhwi_p (DECL_SIZE (SYMBOL_REF_DECL (x))) ? tree_to_uhwi (DECL_SIZE (SYMBOL_REF_DECL (x))) : 2*BITS_PER_WORD); } but I cannot really test it.
On 23 Jun 2024, at 11:47, ebotcazou at gcc dot gnu.org <gcc-bugzilla@gcc.gnu.org> wrote: > > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115591 > > --- Comment #2 from Eric Botcazou <ebotcazou at gcc dot gnu.org> --- > The immediate fix is: > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > index c17141d909a..5e34dc92210 100644 > --- a/gcc/config/riscv/riscv.cc > +++ b/gcc/config/riscv/riscv.cc > @@ -1702,7 +1702,9 @@ riscv_valid_lo_sum_p (enum riscv_symbol_type sym_type, > machine_mode mode, > align = (SYMBOL_REF_DECL (x) > ? DECL_ALIGN (SYMBOL_REF_DECL (x)) > : 1); > - size = (SYMBOL_REF_DECL (x) && DECL_SIZE (SYMBOL_REF_DECL (x)) > + size = (SYMBOL_REF_DECL (x) > + && DECL_SIZE (SYMBOL_REF_DECL (x)) > + && tree_fits_uhwi_p (DECL_SIZE (SYMBOL_REF_DECL (x))) > ? tree_to_uhwi (DECL_SIZE (SYMBOL_REF_DECL (x))) > : 2*BITS_PER_WORD); > } > > but I cannot really test it. Can I help with testing? (beyond applying the patch and checking that the error is fixed)
bug.adb compiles without error after applying the patch.
Jeff, this looks like an obvious fix given the comment just above the modified code, but only a maintainer can probably assess that.
Eric, Just threw this into my tester. Figure ~90 minutes to get back the cross results. I assume that if we go forward that you'll handle putting together a regression test since it's Ada source?
> Just threw this into my tester. Figure ~90 minutes to get back the cross > results. Thanks! > I assume that if we go forward that you'll handle putting together a > regression test since it's Ada source? Yes, will do.
Passed without issue. I'll go ahead and ACK the patch here. It's good to go IMHO.
The master branch has been updated by Eric Botcazou <ebotcazou@gcc.gnu.org>: https://gcc.gnu.org/g:8bc5561c43b195e1638e5acace8b41b3f7512be3 commit r15-1870-g8bc5561c43b195e1638e5acace8b41b3f7512be3 Author: Eric Botcazou <ebotcazou@adacore.com> Date: Sat Jul 6 11:56:19 2024 +0200 RISC-V: fix internal error on global variable-length array This is an ICE in the RISC-V back-end calling tree_to_uhwi on the DECL_SIZE of a global variable-length array. gcc/ PR target/115591 * config/riscv/riscv.cc (riscv_valid_lo_sum_p): Add missing test on tree_fits_uhwi_p before calling tree_to_uhwi. gcc/testsuite/ * gnat.dg/array41.ads, gnat.dg/array41.adb: New test.
Thanks Jeff! Any objection to backporting the fix onto the 14 branch?
No objections at all. Go for it whenever it's convenient for you.
The releases/gcc-14 branch has been updated by Eric Botcazou <ebotcazou@gcc.gnu.org>: https://gcc.gnu.org/g:6c5ef4944e3858e7daf1ef75a3afa93a3ece37c5 commit r14-10385-g6c5ef4944e3858e7daf1ef75a3afa93a3ece37c5 Author: Eric Botcazou <ebotcazou@adacore.com> Date: Sat Jul 6 11:56:19 2024 +0200 RISC-V: fix internal error on global variable-length array This is an ICE in the RISC-V back-end calling tree_to_uhwi on the DECL_SIZE of a global variable-length array. gcc/ PR target/115591 * config/riscv/riscv.cc (riscv_valid_lo_sum_p): Add missing test on tree_fits_uhwi_p before calling tree_to_uhwi. gcc/testsuite/ * gnat.dg/array41.ads, gnat.dg/array41.adb: New test.
Thanks! Fixed on mainline and 14 branch.