|
From: Mark W. <mj...@re...> - 2014-05-16 22:40:42
|
Hi Carl,
The functional support patch (KDE#334834) didn't apply cleanly, but the
changes were minor. Basically there are a couple of lines that say:
-# if defined(VG_PLAT_USES_PPCTOC) || defined(VGP_ppc64le_linux)
+# if defined(VG_PLAT_USES_PPCTOC) || defined(VGP_ppc64le_linux)
The first - line should not have the second || defined part in it.
Then it applies cleanly.
Fedora now has a koji builder for ppc64le (still very experimental) so I
tried to apply your patches and fired off a build. The good news is that
it all builds and the result works for ./vg-in-place /bin/true.
But the tests did not all compile so make regtest didn't run.
I haven't investigated yet. But I am seeing two tests that fail to
build:
gcc -Winline -Wall -Wshadow -g -m64 -Winline -Wall -O -g -mregnames -maltivec -m64 -DHAS_ALTIVEC -Wno-long-long -g -pipe -Wall -Werror=format-security -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -Wwrite-strings -fno-stack-protector -Wno-write-strings -Wl,-z,relro -o jm-insns jm_insns-jm-insns.o
/usr/bin/ld: .opd not allowed in ABI version 2
jm_insns-jm-insns.o: error adding symbols: Bad value
collect2: error: ld returned 1 exit status
gcc -DHAVE_CONFIG_H -I. -I../../.. -I../../.. -I../../../include -I../../../coregrind -I../../../include -I../../../VEX/pub -I../../../VEX/pub -DVGA_ppc64le=1 -DVGO_linux=1 -DVGP_ppc64le_linux=1 -DVGPV_ppc64le_linux_vanilla=1 -Winline -Wall -Wshadow -g -m64 -c -o ldst_multiple.o ldst_multiple.c
ldst_multiple.c: In function 'main':
ldst_multiple.c:180:5: error: frame pointer required, but reserved
int main(void)
^
ldst_multiple.c:31:18: note: for 'r31'
register HWord_t r31 asm("r31");
^
Note that it might just be the build environment (wrong gcc or binutils
version) on that build machine, which is still experimental. I don't
have direct access to the build machine, so debugging will take a bit of
effort.
The full build.log is at:
http://ppc-le.koji.fedoraproject.org/kojifiles/work/tasks/3017/43017/build.log
The srpm and rpms can be found at:
http://ppc-le.koji.fedoraproject.org/koji/taskinfo?taskID=43017
Cheers,
Mark
|
|
From: Maynard J. <may...@us...> - 2014-05-19 12:41:40
|
On 05/16/2014 04:50 PM, Mark Wielaard wrote: > Hi Carl, > > The functional support patch (KDE#334834) didn't apply cleanly, but the > changes were minor. Basically there are a couple of lines that say: > > -# if defined(VG_PLAT_USES_PPCTOC) || defined(VGP_ppc64le_linux) > +# if defined(VG_PLAT_USES_PPCTOC) || defined(VGP_ppc64le_linux) > > The first - line should not have the second || defined part in it. > Then it applies cleanly. > > Fedora now has a koji builder for ppc64le (still very experimental) so I > tried to apply your patches and fired off a build. The good news is that > it all builds and the result works for ./vg-in-place /bin/true. > > But the tests did not all compile so make regtest didn't run. > > I haven't investigated yet. But I am seeing two tests that fail to > build: Mark, Carl, Both of these issues are fixed with the testcase patch available in https://bugs.kde.org/show_bug.cgi?id=334836. -Maynard > gcc -Winline -Wall -Wshadow -g -m64 -Winline -Wall -O -g -mregnames -maltivec -m64 -DHAS_ALTIVEC -Wno-long-long -g -pipe -Wall -Werror=format-security -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -Wwrite-strings -fno-stack-protector -Wno-write-strings -Wl,-z,relro -o jm-insns jm_insns-jm-insns.o > /usr/bin/ld: .opd not allowed in ABI version 2 > jm_insns-jm-insns.o: error adding symbols: Bad value > collect2: error: ld returned 1 exit status > > gcc -DHAVE_CONFIG_H -I. -I../../.. -I../../.. -I../../../include -I../../../coregrind -I../../../include -I../../../VEX/pub -I../../../VEX/pub -DVGA_ppc64le=1 -DVGO_linux=1 -DVGP_ppc64le_linux=1 -DVGPV_ppc64le_linux_vanilla=1 -Winline -Wall -Wshadow -g -m64 -c -o ldst_multiple.o ldst_multiple.c > ldst_multiple.c: In function 'main': > ldst_multiple.c:180:5: error: frame pointer required, but reserved > int main(void) > ^ > ldst_multiple.c:31:18: note: for 'r31' > register HWord_t r31 asm("r31"); > ^ > > Note that it might just be the build environment (wrong gcc or binutils > version) on that build machine, which is still experimental. I don't > have direct access to the build machine, so debugging will take a bit of > effort. > > The full build.log is at: > http://ppc-le.koji.fedoraproject.org/kojifiles/work/tasks/3017/43017/build.log > The srpm and rpms can be found at: > http://ppc-le.koji.fedoraproject.org/koji/taskinfo?taskID=43017 > > Cheers, > > Mark > |