[Bug debug/81936] ICE in dwarf2out_die_ref_for_decl, at dwarf2out.c:5543

["rguenther at suse dot de" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81936

--- Comment #4 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #2)
> How do I reproduce, aka configure gcc?

The ICE is actually in the host compiler, but it doesn't seem to reproduce in a
non-offloading setup.

Host compiler:

OFFLOAD_TARGET_NAMES=nvptx-none
Target: x86_64-none-linux-gnu
Configured with: configure --build=x86_64-linux-gnu --host=x86_64-linux-gnu
--target=x86_64-none-linux-gnu --enable-threads --disable-libmudflap
--disable-libstdcxx-pch --disable-multilib --disable-libsanitizer
--disable-libcc1 --enable-checking=yes,rtl --with-gnu-as --with-gnu-ld
--enable-languages=c,c++,fortran --enable-shared --enable-lto  --disable-nls
--prefix=prefix --enable-offload-targets=nvptx-none --with-sysroot=libc
--with-build-sysroot=libc --with-gmp=usr --with-mpfr=usr --with-mpc=usr
--with-isl=usr --with-cuda-driver-include=cuda-runtime/include
--with-cuda-driver-lib=cuda-driver/lib --enable-libgomp --disable-libitm
--enable-libatomic --disable-libssp --disable-libcc1 --with-python-dir=python
--with-build-time-tools=bin

Offloading compiler:

Target: nvptx-none
Configured with: configure --build=x86_64-linux-gnu --host=x86_64-linux-gnu
--target=nvptx-none --enable-threads --disable-libmudflap
--disable-libstdcxx-pch --disable-multilib --disable-libsanitizer
--disable-libcc1 --enable-checking=yes,rtl --with-gnu-as --with-gnu-ld
--enable-languages=c,c++,fortran --enable-shared --enable-lto --disable-nls
--prefix=prefix --with-headers=yes --enable-languages=c,lto,fortran
--disable-sjlj-exceptions --disable-lto --without-gnu-ld --without-gnu-as
--with-newlib --enable-multilib
--enable-as-accelerator-for=x86_64-none-linux-gnu --with-sysroot=nvptx-none
--with-build-sysroot=nvptx-none --with-gmp=usr --with-mpfr=usr --with-mpc=usr
--with-isl=usr --with-cuda-driver-include=include
--with-cuda-driver-lib=cuda-driver/lib --enable-libgomp --disable-libitm
--disable-libatomic --disable-libssp --disable-libcc1 --with-python-dir=python
--with-build-time-tools=bin

--- Comment #5 from rguenther at suse dot de <rguenther at suse dot de> ---
On Wed, 23 Aug 2017, vries at gcc dot gnu.org wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81936
> 
> --- Comment #3 from Tom de Vries <vries at gcc dot gnu.org> ---
> ICE is due to assert:
> ...
>     at gcc/dwarf2out.c:5542
> 5542      gcc_assert (die->die_tag == DW_TAG_compile_unit
> 5543                  && die->die_id.die_symbol != NULL);
> ...
> 
> The assert triggers because die->die_id.die_symbol is NULL:
> ...
> (gdb) p die->die_tag
> $3 = DW_TAG_compile_unit
> (gdb) p die->die_id.die_symbol
> $4 = 0x0
> ...
> 
> If we go up we see we're trying to find a die_ref for 'expr':
> ...
> (gdb) up
> #7  0x0000000000b9d804 in lto_write_tree_1 (ob=0x2771d30, expr=0x7ffff7521180,
> ref_p=Unhandled dwarf expression opcode 0xf3)
>     at gcc/lto-streamer-out.c:422
> 422               && debug_hooks->die_ref_for_decl (expr, &sym, &off))
> ...
> 
> and expr is:
> ...
> (gdb) call debug_generic_expr (expr)
> .omp_data_i
> ...

The assert says that if we have a DIE at stream-in time then it should
refer to early debug.  What created the DIE for .omp_data_i and/or
what failed to properly set its parent?