[Buildroot] [Bug 8766] New: Compiling host-gcc-final-4.9.3 broken on i386

[Buildroot] [Bug 8766] New: Compiling host-gcc-final-4.9.3 broken on i386

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=8766

            Bug ID: 8766
           Summary: Compiling host-gcc-final-4.9.3 broken on i386
           Product: buildroot
           Version: unspecified
          Hardware: All
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P5
         Component: Other
          Assignee: unassigned at buildroot.uclibc.org
          Reporter: bernd.kuhls at t-online.de
                CC: buildroot at uclibc.org
  Target Milestone: ---

When compiling buildroot git master (c6e3511) with this defconfig

BR2_x86_i386=y

this error occurs:

/home/bernd/buildroot/br4_i386/output/build/host-gcc-final-4.9.3/build/./gcc/xgcc
-B/home/bernd/buildroot/br4_i386/output/build/host-gcc-final-4.9.3/build/./gcc/
-B/home/bernd/buildroot/br4_i386/output/host/usr/i386-buildroot-linux-uclibc/bin/
-B/home/bernd/buildroot/br4_i386/output/host/usr/i386-buildroot-linux-uclibc/lib/
-isystem
/home/bernd/buildroot/br4_i386/output/host/usr/i386-buildroot-linux-uclibc/include
-isystem
/home/bernd/buildroot/br4_i386/output/host/usr/i386-buildroot-linux-uclibc/sys-include
   -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Os -O2 
-D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64  -Os  -DIN_GCC
 -DCROSS_DIRECTORY_STRUCTURE  -W -Wall -Wno-narrowing -Wwrite-strings
-Wcast-qual -Wstrict-prototypes -Wmissing-prototypes -Wold-style-definition 
-isystem ./include   -fpic -mlong-double-80 -g -DIN_LIBGCC2 -fbuilding-libgcc
-fno-stack-protector   -fpic -mlong-double-80 -I. -I. -I../.././gcc
-I../../../libgcc -I../../../libgcc/. -I../../../libgcc/../gcc
-I../../../libgcc/../include    -DUSE_TLS -o _gcov_ior_profiler.o -MT
_gcov_ior_profiler.o -MD -MP -MF _gcov_ior_profiler.dep -DL_gcov_ior_profiler
-c ../../../libgcc/libgcov-profiler.c
../../../libgcc/generic-morestack-thread.c:41:21: fatal error: pthread.h: No
such file or directory
 #include <pthread.h>
                     ^
compilation terminated.
../../../libgcc/shared-object.mk:14: recipe for target
'generic-morestack-thread.o' failed
make[3]: *** [generic-morestack-thread.o] Error 1
make[3]: *** Waiting for unfinished jobs....
make[3]: Leaving directory
'/home/bernd/buildroot/br4_i386/output/build/host-gcc-final-4.9.3/build/i386-buildroot-linux-uclibc/libgcc'
Makefile:10153: recipe for target 'all-target-libgcc' failed
make[2]: *** [all-target-libgcc] Error 2
make[2]: Leaving directory
'/home/bernd/buildroot/br4_i386/output/build/host-gcc-final-4.9.3/build'
Makefile:840: recipe for target 'all' failed
make[1]: *** [all] Error 2
make[1]: Leaving directory
'/home/bernd/buildroot/br4_i386/output/build/host-gcc-final-4.9.3/build'
package/pkg-generic.mk:195: recipe for target
'/home/bernd/buildroot/br4_i386/output/build/host-gcc-final-4.9.3/.stamp_built'
failed
make: ***
[/home/bernd/buildroot/br4_i386/output/build/host-gcc-final-4.9.3/.stamp_built]
Error 2

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 8766] Compiling host-gcc-final-4.9.3 broken on i386

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=8766

--- Comment #1 from Thomas Petazzoni <thomas.petazzoni@free-electrons.com> ---
Thanks Bernd for the report. This is an issue you already reported in the past,
I found your previous report and some related discussion at
https://patchwork.ozlabs.org/patch/484002/.

Basically, as I understand it, the problem is that uClibc-ng only offers the
NPTL thread implementation for x86. But since i386 does not have the necessary
instructions to implement NPTL, we only offer the "none" thread implementation,
which explains why <pthread.h> is missing.

So I think there are really two questions here:

 1/ Is it correct that uClibc-ng provides no thread implementation at all for
i386. This is something to discuss with Waldemar.

 2/ Are we also seeing this <pthread.h> missing problem when building with no
threads on i486+. If we also have this problem, then I don't see how the "none"
thread implementation can still be useful.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 8766] Compiling host-gcc-final-4.9.3 broken on i386

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=8766

--- Comment #2 from Thomas Petazzoni <thomas.petazzoni@free-electrons.com> ---
(In reply to Thomas Petazzoni from comment #1)
>  2/ Are we also seeing this <pthread.h> missing problem when building with no
> threads on i486+. If we also have this problem, then I don't see how the "none" 
> thread implementation can still be useful.

I just tested, it also fails with the same error on i486 with thread support
disabled. I.E this defconfig:

BR2_x86_i486=y
BR2_PTHREADS_NONE=y

fails to build with:

../../../libgcc/generic-morestack-thread.c:41:21: fatal error: pthread.h: No
such file or directory
 #include <pthread.h>
                     ^
compilation terminated.
make[4]: *** [generic-morestack-thread.o] Error 1
make[4]: *** Waiting for unfinished jobs....
make[3]: *** [all-target-libgcc] Error 2
make[2]: *** [all] Error 2
make[1]: ***
[/home/test/outputs/i486-nothread/build/host-gcc-final-4.9.3/.stamp_built]
Error 2
make: *** [_all] Error 2

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 8766] Compiling host-gcc-final-4.9.3 broken on i386

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=8766

--- Comment #3 from Thomas Petazzoni <thomas.petazzoni@free-electrons.com> ---
After some research, it seems to be an x86 specific issue. The file causing
problems is libgcc/generic-morestack-thread.c, which gets included in the build
by libgcc/config/t-stack. It is related to the implementation of -fsplit-stack,
which is only supported on x86 32/64.

So on other architectures, a no thread configuration should still build fine.
However, seeing that on a major arch like x86 the nothread support has been
broken for a while (-fsplit-stack has been around since gcc 4.6) really makes
me wonder if we should continue to support the no thread solution.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 8766] Compiling host-gcc-final-4.9.3 broken on i386

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=8766

--- Comment #4 from Peter Korsgaard <jacmet@uclibc.org> ---
(In reply to Thomas Petazzoni from comment #3)

I also doubt the no-thread option gets used very much. Perhaps the way forward
is to deprecate it and then make threads always available, similar to how we
did for largefile/IPv6?

If i386 support blocks this, then I suggest we drop the i386 variant.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 8766] Compiling host-gcc-final-4.9.3 broken on i386

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=8766

Thomas Petazzoni <thomas.petazzoni@free-electrons.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #5 from Thomas Petazzoni <thomas.petazzoni@free-electrons.com> ---
i386 support has been deprecated as of
https://git.buildroot.net/buildroot/commit/?id=6cb4814c87cb1282b4a1d35e73743e8dd1b12ec5.

x86 non-threaded configuration has been fixed as of
https://git.buildroot.net/buildroot/commit/?id=2631219f648f4aa13fa36d06a483e4507cb6f3d7.

-- 
You are receiving this mail because:
You are on the CC list for the bug.