aarch64 host based sysutils/u-boot-{pine64, rock64, rpi[34]} builds fail for: "aarch64-none-elf-gcc: fatal error: -fuse-linker-plugin, but liblto_plugin.so not found"
Mark Millard
marklmi at yahoo.com
Tue Apr 28 21:29:06 UTC 2020
[Looks like more than objdump may be involved for
/usr/local/aarch64-none-elf/bin/ use.]
On 2020-Apr-28, at 14:00, Mark Millard <marklmi at yahoo.com> wrote:
> [Found a cause of the poudiere vs. not distinction.]
>
> On 2020-Apr-28, at 13:01, Mark Millard <marklmi at yahoo.com> wrote:
>>
>> On 2020-Apr-28, at 09:23, Mark Millard <marklmi at yahoo.com> wrote:
>>
>>> On 2020-Apr-28, at 07:39, Emmanuel Vadot <manu at bidouilliste.com> wrote:
>>>
>>>> On Mon, 27 Apr 2020 20:14:47 -0700
>>>> Mark Millard <marklmi at yahoo.com> wrote:
>>>>
>>>>> On 2020-Apr-27, at 17:15, Mark Millard <marklmi at yahoo.com> wrote:
>>>>>
>>>>>> On 2020-Apr-27, at 11:46, Emmanuel Vadot <manu at bidouilliste.com> wrote:
>>>>>>
>>>>>>> On Mon, 27 Apr 2020 12:32:46 +0200
>>>>>>> Emmanuel Vadot <manu at bidouilliste.com> wrote:
>>>>>>>
>>>>>>>> On Sun, 26 Apr 2020 12:13:46 -0700
>>>>>>>> . . .
>>>>>>>
>>>>>>> This is solved with r533162.
>>>>>>> I've tested that sysutils/u-boot-pine64-lts build and boots fine.
>>>>>>> Let me know if you have still any issues.
>>>>>>
>>>>>> The poudriere-devel runs based on updating /usr/ports content
>>>>>> ended up scheduling 190 package builds, including a couple of
>>>>>> llvm*'s. Depending on where in the sequence things happen, it
>>>>>> could be some time before having such full-native results.
>>>>>>
>>>>>> In a amd64->aarch64 pouriere-devel based cross-build, with
>>>>>> nxb-bin in use (a faster build environment in my context),
>>>>>> I got:
>>>>>>
>>>>>> . . .
>>>>>> gmake[4]: Leaving directory '/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/aarch64-none-elf/ilp32/libgcc'
>>>>>> gmake[3]: Leaving directory '/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/aarch64-none-elf/libgcc'
>>>>>> gmake[2]: Leaving directory '/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/aarch64-none-elf/libgcc'
>>>>>> gmake[1]: Leaving directory '/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build'
>>>>>> ====> Compressing man pages (compress-man)
>>>>>> ===========================================================================
>>>>>> =======================<phase: package >============================
>>>>>> ===> Building package for aarch64-none-elf-gcc-8.4.0_1
>>>>>> pkg-static: Unable to access file /wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/stage/usr/local/lib/gcc/aarch64-none-elf/8.4.0/plugin/gtype.state:No such file or directory
>>>>>> pkg-static: Unable to access file /wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/stage/usr/local/lib/gcc/aarch64-none-elf/8.4.0/plugin/include/ada/gcc-interface/ada-tree.def:No such file or direct
>>>>>> ory
>>>>>> pkg-static: Unable to access file /wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/stage/usr/local/lib/gcc/aarch64-none-elf/8.4.0/plugin/include/addresses.h:No such file or directory
>>>>>> pkg-static: Unable to access file /wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/stage/usr/local/lib/gcc/aarch64-none-elf/8.4.0/plugin/include/alias.h:No such file or directory
>>>>>> . . . (long list) . . .
>>>>>>
>>>>>>
>>>>>> That leads to:
>>>>>>
>>>>>> [00:29:06] [04] [00:10:01] Saved devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 wrkdir to: /usr/local/poudriere/data/wrkdirs/FBSDFSSDjailCortexA53-default/default/aarch64-none-elf-gcc-8.4.0_1.tar
>>>>>> [00:29:10] [04] [00:10:05] Finished devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1: Failed: package
>>>>>> [00:29:12] [04] [00:10:07] Skipping sysutils/atf-rk3328 | atf-rk3328-v2.3: Dependent port devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 failed
>>>>>> [00:29:12] [04] [00:10:07] Skipping sysutils/atf-sun50i_a64 | atf-sun50i_a64-v2.3: Dependent port devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 failed
>>>>>> [00:29:12] [04] [00:10:07] Skipping sysutils/u-boot-pine64 | u-boot-pine64-2020.04: Dependent port devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 failed
>>>>>> [00:29:12] [04] [00:10:07] Skipping sysutils/u-boot-rock64 | u-boot-rock64-2020.04: Dependent port devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 failed
>>>>>> [00:29:12] [04] [00:10:07] Skipping sysutils/u-boot-rpi3 | u-boot-rpi3-2020.04: Dependent port devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 failed
>>>>>> [00:29:12] [04] [00:10:07] Skipping sysutils/u-boot-rpi4 | u-boot-rpi4-2020.04: Dependent port devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 failed
>>>>>>
>>>>>>
>>>>>>
>>>>>> I'm guessing that the full-native builds will do similarly.
>>>>>> But we will eventually see.
>>>>>
>>>>> One of the aarch64 environments finally got there:
>>>>>
>>>>> [07:40:23] [02] [00:23:04] Saved devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 wrkdir to: /usr/local/poudriere/data/wrkdirs/FBSDFSSDjailCortexA57-default/default/aarch64-none-elf-gcc-8.4.0_1.tar
>>>>> [07:40:42] [02] [00:23:23] Finished devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1: Failed: package
>>>>> [07:40:55] [02] [00:23:36] Skipping sysutils/u-boot-pine64 | u-boot-pine64-2020.04: Dependent port devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 failed
>>>>> [07:40:55] [02] [00:23:36] Skipping sysutils/u-boot-rock64 | u-boot-rock64-2020.04: Dependent port devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 failed
>>>>> [07:40:55] [02] [00:23:36] Skipping sysutils/u-boot-rpi3 | u-boot-rpi3-2020.04: Dependent port devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 failed
>>>>> [07:40:55] [02] [00:23:36] Skipping sysutils/u-boot-rpi4 | u-boot-rpi4-2020.04: Dependent port devel/aarch64-none-elf-gcc | aarch64-none-elf-gcc-8.4.0_1 failed
>>>>>
>>>>> It failed for the same sorts of reasons as the cross build:
>>>>>
>>>>> =======================<phase: package >============================
>>>>> ===> Building package for aarch64-none-elf-gcc-8.4.0_1
>>>>> pkg-static: Unable to access file /wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/stage/usr/local/lib/gcc/aarch64-none-elf/8.4.0/plugin/gtype.state:No such file or directory
>>>>> pkg-static: Unable to access file /wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/stage/usr/local/lib/gcc/aarch64-none-elf/8.4.0/plugin/include/ada/gcc-interface/ada-tree.def:No such file or direct
>>>>> ory
>>>>> . . .
>>>>>
>>>>
>>>> I can reproduce that with poudriere too.
>>>> For some reason this doesn't happened when doing a make package ...
>>>
>>> It also happened for the poudriere-based build of
>>> devel/arm-none-eabi-gcc :
>>>
>>> [00:26:15] [07] [00:13:23] Saving devel/arm-none-eabi-gcc | arm-none-eabi-gcc-8.4.0_1 wrkdir
>>> [00:26:23] [07] [00:13:31] Saved devel/arm-none-eabi-gcc | arm-none-eabi-gcc-8.4.0_1 wrkdir to: /usr/local/poudriere/data/wrkdirs/FBSDFSSDjailCortexA53-default/default/arm-none-eabi-gcc-8.4.0_1.tar
>>> [00:26:28] [07] [00:13:36] Finished devel/arm-none-eabi-gcc | arm-none-eabi-gcc-8.4.0_1: Failed: package
>>> [00:26:30] [07] [00:13:38] Skipping sysutils/u-boot-orangepi-plus-2e | u-boot-orangepi-plus-2e-2020.04: Dependent port devel/arm-none-eabi-gcc | arm-none-eabi-gcc-8.4.0_1 failed
>>> [00:26:30] [07] [00:13:38] Skipping sysutils/u-boot-rpi2 | u-boot-rpi2-2020.04: Dependent port devel/arm-none-eabi-gcc | arm-none-eabi-gcc-8.4.0_1 failed
>>> [00:26:30] [07] [00:13:38] Skipping sysutils/u-boot-sinovoip-bpi-m3 | u-boot-sinovoip-bpi-m3-2020.04: Dependent port devel/arm-none-eabi-gcc | arm-none-eabi-gcc-8.4.0_1 failed
>>>
>>> because of:
>>>
>>> =======================<phase: package >============================
>>> ===> Building package for arm-none-eabi-gcc-8.4.0_1
>>> pkg-static: Unable to access file /wrkdirs/usr/ports/devel/arm-none-eabi-gcc/work/stage/usr/local/lib/gcc/arm-none-eabi/8.4.0/plugin/gtype.state:No such file or directory
>>> pkg-static: Unable to access file /wrkdirs/usr/ports/devel/arm-none-eabi-gcc/work/stage/usr/local/lib/gcc/arm-none-eabi/8.4.0/plugin/include/ada/gcc-interface/ada-tree.def:No such file or directory
>>> pkg-static: Unable to access file /wrkdirs/usr/ports/devel/arm-none-eabi-gcc/work/stage/usr/local/lib/gcc/arm-none-eabi/8.4.0/plugin/include/addresses.h:No such file or directory
>>> pkg-static: Unable to access file /wrkdirs/usr/ports/devel/arm-none-eabi-gcc/work/stage/usr/local/lib/gcc/arm-none-eabi/8.4.0/plugin/include/alias.h:No such file or directory
>>> pkg-static: Unable to access file /wrkdirs/usr/ports/devel/arm-none-eabi-gcc/work/stage/usr/local/lib/gcc/arm-none-eabi/8.4.0/plugin/include/all-tree.def:No such file or directory
>>> . . .
>>>
>>
>> Some other differences between a poudriere based build
>> and a portmaster based build for the aarch64 case follow.
>> The paths with /mnt/prefixes are to the portmaster based
>> build:
>>
>> # diff -r /wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/ /mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/ | grep -i "enable_plugin"
>> < enable_plugin = no
>>> enable_plugin = yes
>> < /* #undef ENABLE_PLUGIN */
>>> #define ENABLE_PLUGIN 1
>> < enable_plugin='no'
>>> enable_plugin='yes'
>>> #define ENABLE_PLUGIN 1
>> < S["enable_plugin"]="no"
>>> S["enable_plugin"]="yes"
>>> D["ENABLE_PLUGIN"]=" 1"
>>
>> While looking for differences that might contribute to the
>> "no"s above, I ran into something possibly unrelated but
>> important in the future . . .
>>
>> One thing is that I have a WITHOUT_BINUTILS= system build as
>> part of testing that such will work when if fully goes away.
>> I see differences in part based on that:
>>
>> 1383c1382,1383
>> < configure:8340: result: no
>> ---
>>> configure:8326: found /usr/local/bin/as
>>> configure:8337: result: as
>> 1407c1407,1408
>> < configure:9740: result: no
>> ---
>>> configure:9726: found /usr/local/bin/objdump
>>> configure:9737: result: objdump
>> . . .
>> 2369a2371
>>> ac_cv_prog_AS=as
>> 2377a2380
>>> ac_cv_prog_OBJDUMP=objdump
>>
>> (No evidence so far that such contributes to lack of
>> plugins being enabled. It is just a difference that
>> I ran into while looking.)
>>
>> The devel/*-none-*-gcc examples do not cause
>> /usr/local/bin/ instances of as or objdump to
>> exist in poudriere-devel contexts, even when
>> the system does not have the commands of itself.
>> At some point this likely will need to change.
>>
>>
>>
>> As I have time I'll look around more at the build
>> materials and see if I find an explanation for
>> plugins not being enabled under poudriere-devel
>> based builds.
>
> Turns out that objdump is important and makes the
> distinction in my context . . .
>
> work/gcc-8.4.0/config/gcc-plugin.m4 has:
>
> . . .
> AC_DEFUN([GCC_ENABLE_PLUGINS],
> [# Check for plugin support
> AC_ARG_ENABLE(plugin,
> [AS_HELP_STRING([--enable-plugin], [enable plugin support])],
> enable_plugin=$enableval,
> enable_plugin=yes; default_plugin=yes)
>
> pluginlibs=
> plugin_check=yes
>
> case "${host}" in
> . . .
> *)
> if test x$build = x$host; then
> export_sym_check="objdump${exeext} -T"
> elif test x$host = x$target; then
> export_sym_check="$gcc_cv_objdump -T"
> else
> export_sym_check=
> fi
> ;;
> esac
> . . .
> if test x"$enable_plugin" = x"yes" -a x"$plugin_check" = x"yes"; then
>
> AC_MSG_CHECKING([for exported symbols])
> if test "x$export_sym_check" != x; then
> echo "int main() {return 0;} int foobar() {return 0;}" > conftest.c
> ${CC} ${CFLAGS} ${LDFLAGS} conftest.c -o conftest$ac_exeext > /dev/null 2>&1
> if $export_sym_check conftest$ac_exeext | grep foobar > /dev/null; then
> : # No need to use a flag
> AC_MSG_RESULT([yes])
> else
> AC_MSG_RESULT([yes])
> AC_MSG_CHECKING([for -rdynamic])
> ${CC} ${CFLAGS} ${LDFLAGS} -rdynamic conftest.c -o conftest$ac_exeext > /dev/null 2>&1
> if $export_sym_check conftest$ac_exeext | grep foobar > /dev/null; then
> plugin_rdynamic=yes
> pluginlibs="-rdynamic"
> else
> plugin_rdynamic=no
> enable_plugin=no
> fi
> AC_MSG_RESULT([$plugin_rdynamic])
> fi
> else
> AC_MSG_RESULT([unable to check])
> fi
> . . .
>
> But (both the poudriere-devel based build and the portmaster
> build show such so I picked to show just one):
>
> /wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_objdump=/usr/local/aarch64-none-elf/bin/objdump
> /wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_objdump='/usr/local/aarch64-none-elf/bin/objdump'
> /wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.cache:gcc_cv_objdump=${gcc_cv_objdump=/usr/local/aarch64-none-elf/bin/objdump}
>
> Under poudiere-devel there is no /usr/local/aarch64-none-elf/bin/objdump path.
> Under make/portmaster there may be an old /usr/local/aarch64-none-elf/bin/objdump path.
>
> In my environments (showing one example context):
>
> # ls -ldT /usr/local/aarch64-none-elf/bin/objdump
> -r-xr-xr-x 2 root wheel 8593272 Jan 30 01:05:50 2020 /usr/local/aarch64-none-elf/bin/objdump
>
> So in portmaster (or make) the old /usr/local/aarch64-none-elf/bin/objdump
> is used but in poudriere the path used fails and that failure is not
> handled distinctly as "objdump tool missing".
>
> It looks like devel/aarch64-none-elf-gcc/Makefile would need
> changes for WITHOUT_BINUTILS= contexts in order to make sure
> that an objdump exists via a known/used path, including in
> poudriere based builds.
looks like nm, objdump, and readelf have that /usr/local/aarch64-none-elf/bin/
style path issue shown in work/.build/gcc/config.log . There are some ld
references as well.
(Again: /mnt/ is for portmaster build context; otherwise it is poudriere based.)
# grep -B1 /usr/local/aarch64-none-elf/bin/ /wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log /mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-configure:22232: checking what nm to use
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:configure:22240: result: /usr/local/aarch64-none-elf/bin/nm
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-configure:22312: checking what objdump to use
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:configure:22322: result: /usr/local/aarch64-none-elf/bin/objdump
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-configure:22384: checking what readelf to use
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:configure:22394: result: /usr/local/aarch64-none-elf/bin/readelf
--
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-gcc_cv_mkdir_takes_one_arg=no
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_nm=/usr/local/aarch64-none-elf/bin/nm
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-gcc_cv_no_pie=yes
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_objdump=/usr/local/aarch64-none-elf/bin/objdump
--
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-gcc_cv_prog_makeinfo_modern=yes
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_readelf=/usr/local/aarch64-none-elf/bin/readelf
--
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-ORIGINAL_LD_GOLD_FOR_TARGET='/usr/local/bin/aarch64-none-elf-ld.gold'
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:ORIGINAL_NM_FOR_TARGET='/usr/local/aarch64-none-elf/bin/nm'
--
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-gcc_cv_ld='/usr/local/bin/aarch64-none-elf-ld'
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_nm='/usr/local/aarch64-none-elf/bin/nm'
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_objdump='/usr/local/aarch64-none-elf/bin/objdump'
/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_readelf='/usr/local/aarch64-none-elf/bin/readelf'
--
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-configure:22232: checking what nm to use
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:configure:22240: result: /usr/local/aarch64-none-elf/bin/nm
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-configure:22312: checking what objdump to use
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:configure:22322: result: /usr/local/aarch64-none-elf/bin/objdump
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-configure:22384: checking what readelf to use
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:configure:22394: result: /usr/local/aarch64-none-elf/bin/readelf
--
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-gcc_cv_mkdir_takes_one_arg=no
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_nm=/usr/local/aarch64-none-elf/bin/nm
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-gcc_cv_no_pie=yes
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_objdump=/usr/local/aarch64-none-elf/bin/objdump
--
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-gcc_cv_prog_makeinfo_modern=yes
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_readelf=/usr/local/aarch64-none-elf/bin/readelf
--
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-ORIGINAL_LD_GOLD_FOR_TARGET='/usr/local/bin/aarch64-none-elf-ld.gold'
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:ORIGINAL_NM_FOR_TARGET='/usr/local/aarch64-none-elf/bin/nm'
--
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log-gcc_cv_ld='/usr/local/bin/aarch64-none-elf-ld'
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_nm='/usr/local/aarch64-none-elf/bin/nm'
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_objdump='/usr/local/aarch64-none-elf/bin/objdump'
/mnt/wrkdirs/usr/ports/devel/aarch64-none-elf-gcc/work/.build/gcc/config.log:gcc_cv_readelf='/usr/local/aarch64-none-elf/bin/readelf'
===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)
More information about the freebsd-ports
mailing list