[Bug target/96233] New: XGCC-10.1.0 for Target RX: internal compiler error: in dwarf2out_frame_debug_adjust_cfa, at dwarf2cfi.c:1189
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96233 Bug ID: 96233 Summary: XGCC-10.1.0 for Target RX: internal compiler error: in dwarf2out_frame_debug_adjust_cfa, at dwarf2cfi.c:1189 Product: gcc Version: 10.1.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: target Assignee: unassigned at gcc dot gnu.org Reporter: hdu...@tangerine-soft.de Target Milestone: --- Created attachment 4 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=4=edit Code snippet that causes the error ICE happened in a trivial c-source code with main() entry (see attachment) I got following errors. $ gcc-10.1.0-rx-none-elf/bin/rx-none-elf-gcc -g -o main.o -c ./main.c during RTL pass: dwarf2 ./main.c: In function 'int main(int, char**)': ./main.c:6:1: internal compiler error: in dwarf2out_frame_debug_adjust_cfa, at dwarf2cfi.c:1189 6 | } | ^ libbacktrace could not find executable to open Please submit a full bug report, with preprocessed source if appropriate. See <https://gcc.gnu.org/bugs/> for instructions. However either one of the followings works fine: $ gcc-10.1.0-rx-none-elf/bin/rx-none-elf-gcc -g0 -o main.o -c ./main.c $ gcc-10.1.0-rx-none-elf/bin/rx-none-elf-gcc -g -O1 -o main.o -c ./main.c $ gcc-10.1.0-rx-none-elf/bin/rx-none-elf-gcc -g -O2 -o main.o -c ./main.c $ gcc-10.1.0-rx-none-elf/bin/rx-none-elf-gcc -g -O3 -o main.o -c ./main.c -- I have learned that this error even happens with the rx cross compiler for GCC 9.2.0: -- $ gcc-9.2.0-rx-none-elf/bin/rx-none-elf-gcc -g -o main.o -c ./main.c during RTL pass: dwarf2 ./main.c: In function 'int main(int, char**)': ./main.c:6:1: internal compiler error: in dwarf2out_frame_debug_adjust_cfa, at dwarf2cfi.c:1183 6 | } | ^ libbacktrace could not find executable to open Please submit a full bug report, with preprocessed source if appropriate. See <https://gcc.gnu.org/bugs/> for instructions. -- ...and for an GCC rx cross compiler for GCC 8.2.0 as well: -- $ gcc-8.2.0-rx-none-elf/bin/rx-none-elf-gcc -g -o main.o -c ./main.c during RTL pass: dwarf2 ./main.c: In function 'int main(int, char**)': ./main.c:6:1: internal compiler error: in dwarf2out_frame_debug_adjust_cfa, at dwarf2cfi.c:1171 } ^ libbacktrace could not find executable to open Please submit a full bug report, with preprocessed source if appropriate. See <https://gcc.gnu.org/bugs/> for instructions. If I hand over -g0 instead of -g or if I use an Optimization flag with an level of greater than 0 than the ICE does not happen any longer (verfied for rx-none-elf-gcc-10.1, rx-none-elf-gcc-9.2 and rx-none-elf-gcc-8.2.
[Bug c/89515] New: m32c ICE error: in leaf_function_p, at final.c:4492 when compiling newlib 3.1.0 with XGCC 8.3.0
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89515 Bug ID: 89515 Summary: m32c ICE error: in leaf_function_p, at final.c:4492 when compiling newlib 3.1.0 with XGCC 8.3.0 Product: gcc Version: 8.3.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: c Assignee: unassigned at gcc dot gnu.org Reporter: hdu...@tangerine-soft.de Target Milestone: --- m32c-none-elf-gcc -B/tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/newlib/ -isystem /tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/newlib/targ-include -isystem /tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/include -B/tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/libgloss/m32c -L/tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/libgloss/libnosys -L/tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/libgloss/m32c -mcpu=m32cm -DPACKAGE_NAME=\"newlib\" -DPACKAGE_TARNAME=\"newlib\" -DPACKAGE_VERSION=\"3.1.0\" -DPACKAGE_STRING=\"newlib\ 3.1.0\" -DPACKAGE_BUGREPORT=\"\" -DPACKAGE_URL=\"\" -I. -I/tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/argz -Os -fno-builtin -DPREFER_SIZE_OVER_SPEED -DSMALL_MEMORY -D__NO_SYSCALLS__ -DMISSING_SYSCALL_NAMES -DABORT_PROVIDED -DHAVE_INIT_FINI -g -O2 -mcpu=m32cm -c -o lib_a-argz_create_sep.o `test -f 'argz_create_sep.c' || echo '/tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/argz/'`argz_create_sep.c m32c-none-elf-gcc -B/tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/newlib/ -isystem /tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/newlib/targ-include -isystem /tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/include -B/tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/libgloss/m32c -L/tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/libgloss/libnosys -L/tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/libgloss/m32c -mcpu=m32cm -DPACKAGE_NAME=\"newlib\" -DPACKAGE_TARNAME=\"newlib\" -DPACKAGE_VERSION=\"3.1.0\" -DPACKAGE_STRING=\"newlib\ 3.1.0\" -DPACKAGE_BUGREPORT=\"\" -DPACKAGE_URL=\"\" -I. -I/tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/argz -Os -fno-builtin -DPREFER_SIZE_OVER_SPEED -DSMALL_MEMORY -D__NO_SYSCALLS__ -DMISSING_SYSCALL_NAMES -DABORT_PROVIDED -DHAVE_INIT_FINI -g -O2 -mcpu=m32cm -c -o lib_a-argz_delete.o `test -f 'argz_delete.c' || echo '/tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/argz/'`argz_delete.c m32c-none-elf-gcc -B/tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/newlib/ -isystem /tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/newlib/targ-include -isystem /tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/include -B/tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/libgloss/m32c -L/tmp/build_cross_gcc/newlib-3.1.0_build_m32c-none-elf/m32c-none-elf/m32cm/libgloss/libnosys -L/tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/libgloss/m32c -mcpu=m32cm -DPACKAGE_NAME=\"newlib\" -DPACKAGE_TARNAME=\"newlib\" -DPACKAGE_VERSION=\"3.1.0\" -DPACKAGE_STRING=\"newlib\ 3.1.0\" -DPACKAGE_BUGREPORT=\"\" -DPACKAGE_URL=\"\" -I. -I/tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/argz -Os -fno-builtin -DPREFER_SIZE_OVER_SPEED -DSMALL_MEMORY -D__NO_SYSCALLS__ -DMISSING_SYSCALL_NAMES -DABORT_PROVIDED -DHAVE_INIT_FINI -g -O2 -mcpu=m32cm -c -o lib_a-argz_extract.o `test -f 'argz_extract.c' || echo '/tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/argz/'`argz_extract.c during RTL pass: pro_and_epilogue /tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/argz/argz_count.c: In function 'argz_count': /tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/argz/argz_count.c:25:1: internal compiler error: in leaf_function_p, at final.c:4492 } ^ libbacktrace could not find executable to open Please submit a full bug report, with preprocessed source if appropriate. See <https://gcc.gnu.org/bugs/> for instructions. make[8]: *** [lib_a-argz_count.o] Error 1 make[8]: *** Waiting for unfinished jobs during RTL pass: pro_and_epilogue /tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/argz/argz_append.c: In function 'argz_append': /tmp/build_cross_gcc/newlib-3.1.0_arch/newlib-3.1.0/newlib/libc/argz/argz_append.c:31:1: internal compiler error: in leaf_function_p, at final.c:4492 } ^ libbacktrace could not find executable to open Please submit a full bug report, with preprocessed source if appropriate. See <https://gcc.gnu.org/bugs/> for instructions
[Bug target/89484] ICE in dwarf2out_frame_debug_adjust_cfa, at dwarf2cfi.c:1171 when using __attribute__((interrupt)) w. Debug Info
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89484
--- Comment #4 from Hans-Peter Dusel ---
The Release 8.2.0 of GCC is also affected by this issue.
Here is a snapshot (same test program as mentioned above)
- begin --
alvis:tmp hdusel$ /opt/gcc-8.2.0-rx-none-elf/bin/rx-none-elf-gcc ./test.cpp -c
-g
during RTL pass: dwarf2
./test.cpp: In function 'void ISR_handleUartReceiveData()':
./test.cpp:2:2: internal compiler error: in dwarf2out_frame_debug_adjust_cfa,
at dwarf2cfi.c:1171
{}
^
libbacktrace could not find executable to open
Please submit a full bug report,
with preprocessed source if appropriate.
See
[Bug target/89484] ICE in dwarf2out_frame_debug_adjust_cfa, at dwarf2cfi.c:1171 when using __attribute__((interrupt)) w. Debug Info
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89484 --- Comment #3 from Hans-Peter Dusel --- This issue is not related to the c++ compiler g++. It happens also if one uses gcc.
[Bug c++/89484] Compiler ICE: internal compiler error: in dwarf2out_frame_debug_adjust_cfa, at dwarf2cfi.c:1171 when using __attribute__((interrupt))
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89484 --- Comment #2 from Hans-Peter Dusel --- thor:build hdusel$ /opt/gcc-8.3.0-rx-none-elf/bin/rx-none-elf-g++ -v Using built-in specs. COLLECT_GCC=/opt/gcc-8.3.0-rx-none-elf/bin/rx-none-elf-g++ COLLECT_LTO_WRAPPER=/opt/gcc-8.3.0-rx-none-elf/libexec/gcc/rx-none-elf/8.3.0/lto-wrapper Target: rx-none-elf Configured with: /tmp/build_cross_gcc/gcc-8.3.0_arch/gcc-8.3.0/configure --target=rx-none-elf --prefix=/opt/gcc-8.3.0-rx-none-elf --disable-nls --disable-shared --disable-threads --disable-libssp --enable-multilib --enable-lto --without-headers --with-newlib --with-gnu-as --with-gnu-ld --with-gcc --enable-languages=c,c++ --disable-werror Thread model: single gcc version 8.3.0 (GCC)
[Bug c++/89484] Compiler ICE: internal compiler error: in dwarf2out_frame_debug_adjust_cfa, at dwarf2cfi.c:1171 when using __attribute__((interrupt))
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89484 --- Comment #1 from Hans-Peter Dusel --- Tried to verify this with a GCC 8.3 Crosscompiler (same environment as described) but now for Target ARM. This Crosscompiler together with the mentioned Code-snippet and command line __does not__ cause an ICE.
[Bug c++/89484] New: Compiler ICE: internal compiler error: in dwarf2out_frame_debug_adjust_cfa, at dwarf2cfi.c:1171 when using __attribute__((interrupt))
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89484
Bug ID: 89484
Summary: Compiler ICE: internal compiler error: in
dwarf2out_frame_debug_adjust_cfa, at dwarf2cfi.c:1171
when using __attribute__((interrupt))
Product: gcc
Version: 8.3.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: hdu...@tangerine-soft.de
Target Milestone: ---
Environmant
===
Given is a Crosscompiler GCC 8.3 (official Release)
This Crosscompiler hass been build with Xcode Clang
(Apple LLVM version 10.0.0 (clang-1000.11.45.5) Target:
x86_64-apple-darwin17.7.0)
xGCC Target architecture is "RX" (Renesas RX)
Running Host is: Mac OS X 10.11.6
Problem:
===
If one tries to compile the code as follows:
code start -
static void __attribute__((interrupt)) ISR_handleUartReceiveData(void)
{}
- code end --
...by using the commend line...
$> rx-none-elf-c++ ../Test.cpp -c -g
...The following ICE will happen:
during RTL pass: dwarf2
../Test.cpp: In function 'void ISR_handleUartReceiveData()':
../Test.cpp:6:1: internal compiler error: in dwarf2out_frame_debug_adjust_cfa,
at dwarf2cfi.c:1171
}
^
libbacktrace could not find executable to open
Please submit a full bug report,
with preprocessed source if appropriate.
See <https://gcc.gnu.org/bugs/> for instructions.
Side note:
=
I Discovered that the command line option '-g' __as_well__ the occurence of
'__attribute__((interrupt))' in the program code is vital to trigger this ICE!
[Bug c++/71372] [6/7 Regression] C++ FE drops TREE_THIS_VOLATILE in cp_fold on all tcc_reference trees
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71372 --- Comment #16 from Hans-Peter Dusel <hdu...@tangerine-soft.de> --- Created attachment 38634 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38634=edit Patch for GCC 6.1.0 Patch that applies the changes for official Release of GCC 6.1.0
[Bug c++/71372] [6/7 Regression] C++ FE drops TREE_THIS_VOLATILE in cp_fold on all tcc_reference trees
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71372
--- Comment #15 from Hans-Peter Dusel <hdu...@tangerine-soft.de> ---
(In reply to Jakub Jelinek from comment #12)
> Created attachment 38623 [details]
> gcc7-pr71372.patch
>
> Patch I'm going to bootstrap/regtest.
Thanks guys!
I have applied this patch to official Release 6.1.0 too and want to confirm
that it will work for this too!
Now, the my example code will be compiled to correct code:
Thanks a lot! :-)
P.S: I'll attach the patch for gcc 6.1.0
+++
// testmodule.cpp
void fpoke ()
{
volatile unsigned short* FLASH = (volatile unsigned short*)0x2000;
*(FLASH + 0) = 0xAA;
*(FLASH + 1) = 0x55;
*(FLASH + 0) = 0xA0;
}
hdusel@Merlin /cygdrive/w/tmp
$ /opt/gcc-6.1.0-myos/bin/m68k-elf-g++ -O3 testmodule.cpp -c
hdusel@Merlin /cygdrive/w/tmp
$ /opt/gcc-6.1.0-myos/m68k-elf/bin/objdump.exe -d ./testmodule.o
./testmodule.o: file format elf32-m68k
Disassembly of section .text:
<_Z5fpokePvsii>:
0: 4e56 linkw %fp,#0
4: 207c 8000 moveal #-2147483648,%a0
a: 4e5eunlk %fp
c: 7055moveq #85,%d0
e: 30bc 00aa movew #170,%a0@
12: 33c0 8000 0002 movew %d0,8002 <_Z5fpokePvsii+0x8002>
18: 30bc 00a0 movew #160,%a0@
1c: 4e75rts
[Bug c++/71372] New: GCC Optimization "tree-dse" does not respect access to volatile* upon explicit cast.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71372
Bug ID: 71372
Summary: GCC Optimization "tree-dse" does not respect access to
volatile* upon explicit cast.
Product: gcc
Version: 6.1.0
Status: UNCONFIRMED
Severity: major
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: hdu...@tangerine-soft.de
Target Milestone: ---
GCC 6.1. seems to have an issue by using access to storage which __exlicitly__
is casted to an "volatile*" in conjunction with optimization flag –ftree-dse
This issue has been observed with a crosscompiler which is based on the GCC
6.1.0, host: cygwin (32 bit),
for target architecture m68k.
In order to illustrate the issue see 3 Cases (see details below):
1) Code snippet __without__ any explicit type cast to (volatile*).
GCC 6.1.0 with this code produces correct code.
2) Code Snippet __with__ explicit typecast to “volatile short*”.
In conjunction with optimization level -O1/2/3/s this creates wrong code.
The code is wrong that despite the memory location is marked to be volatile*
the
very first access does not happen and seems to be optimized away!
3) Code Snippet __with__ explicit typecast to “volatile short*”.
In conjunction with optimization level -O1/2/3/s but __excluding__
optimization “tree-dse”.
Now GCC 6.1.0 generates right code.
+++
+ Example 1) ) Code snippet __without__ any explicit type cast.
+ GCC 6.1.0 with this code produces correct code.
+++
// testmodule.cpp
void fpoke (void *addr, short val, int times, int wait)
{
volatile unsigned short* FLASH = (volatile unsigned short*)0x2000;
*(FLASH + 0) = 0xAA;
*(FLASH + 1) = 0x55;
*(FLASH + 0) = 0xA0;
}
hdusel@Merlin /cygdrive/w/tmp
$ /opt/gcc-6.1.0-myos/bin/m68k-elf-g++ -O3 testmodule.cpp -c
hdusel@Merlin /cygdrive/w/tmp
$ /opt/gcc-6.1.0-myos/m68k-elf/bin/objdump.exe -d ./testmodule.o
./testmodule.o: file format elf32-m68k
Disassembly of section .text:
<_Z5fpokePvsii>:
0: 4e56 linkw %fp,#0
4: 207c 2000 moveal #536870912,%a0
a: 4e5eunlk %fp
c: 7055moveq #85,%d0
e: 30bc 00aa movew #170,%a0@
12: 33c0 2000 0002 movew %d0,2002 <_Z5fpokePvsii+0x2002>
18: 30bc 00a0 movew #160,%a0@
1c: 4e75rts
+++
+ Example 2) Code Snippet __with__ explicit typecast to “volatile short*”.
+ In conjunction with optimization level -O1/2/3/s this creates __wrong code__.
+
+ The __very first__ access to the storage address $2000 is missing!
+ (see remarks in the code snippet, below)
+++
// testmodule.cpp
void fpoke (void *addr, short val, int times, int wait)
{
volatile unsigned short* FLASH = (volatile unsigned short*)0x2000;
*(volatile unsigned short*)(FLASH + 0) = 0xAA; // this line will be optimized
away! Why?
*(volatile unsigned short*)(FLASH + 1) = 0x55;
*(volatile unsigned short*)(FLASH + 0) = 0xA0;
}
hdusel@Merlin /cygdrive/w/tmp
$ /opt/gcc-6.1.0-myos/bin/m68k-elf-g++ -O3 testmodule.cpp -c
hdusel@Merlin /cygdrive/w/tmp
$ /opt/gcc-6.1.0-myos/m68k-elf/bin/objdump.exe -d ./testmodule.o
./testmodule.o: file format elf32-m68k
Disassembly of section .text:
<_Z5fpokePvsii>:
0: 4e56 linkw %fp,#0
4: 7055moveq #85,%d0
6: 4e5eunlk %fp
8: 33c0 2000 0002 movew %d0,2002 <_Z5fpokePvsii+0x2002>
e: 303c 00a0 movew #160,%d0
12: 33c0 2000 movew %d0,2000 <_Z5fpokePvsii+0x2000>
18: 4e75rts
+++
+ Example 3) Code Snippet __with__ explicit typecast to “volatile short*”.
+ In conjunction with optimization level -O1/2/3/s but __excluding__
optimization “tree-dse”.
+ Now GCC 6.1.0 creates right code.
+++
// testmodule.cpp (The same as on Issue 2)
void fpoke (void *addr, short val, int times, int wait)
{
volatile unsigned short* FLASH = (volatile unsigned short*)0x2000;
*(volatile unsigned short*)(FLASH + 0) = 0xAA;
*(volatile unsigned short*)(FLASH + 1) = 0x55;
*(volatile unsigned short*)(FLASH + 0) = 0xA0;
}
hdusel@Merlin /cygdrive/w/tmp
$ /opt/gcc-6.1.0-myos/bin/m68k-elf-g++ -O3 -fno-tree-dse testmodule.cpp -c
hdusel@Merlin /cygdrive/w/tmp
$ /opt/gcc-6.1.0-myos/m68k-elf/bin/objdump.exe -d ./testmodule.o
./testmodule.o: file format elf32-m68k
Disassembly of section .text:
