Re: ports head -r487783: on armv7 x11/pixman fails to build: /usr/bin/ld: error: can't create dynamic relocation R_ARM_V4BX against local symbol in readonly segment; recompile object files with -fPIC

Bryan Drewery <[email protected]> Wed, 14 Nov 2018 17:45:25 -0800
Newsgroups	gmane.os.freebsd.devel.x11,gmane.os.freebsd.devel.ports
Organization	FreeBSD
Message-ID	<[email protected]>
On 11/14/18 5:24 PM, Mark Millard wrote:
> [Looks like there are 2 stages involved in that
> 2mmjnk.txt file that I generated. Before
> MACHINE_ARCH is explicitly assigned and after.]
> 
> On 2018-Nov-14, at 15:40, Mark Millard <marklmi at yahoo.com> wrote:
> 
>> [Evidence from inside poudriere bulk -j... -i ports-mgmt/pkg .
>> Use of native /nxb-bin/. . . leads to MACHINE_ARCH being amd64
>> instead of armv7 or the like. See later supporting material.]
>>
>> On 2018-Nov-14, at 14:38, Bryan Drewery <[email protected]> wrote:
>>
>>> On 11/14/18 2:35 PM, Mark Millard wrote:
>>>> [Looking at package fallout logs: the official armv6 and armv7
>>>> builds are using -O2 because of MACHINE_ARCH being defined
>>>> because of qemu-user-static use. (mips too?) The logic in
>>>> share/mk/sys.mk is not causing -O . An implication
>>>> is that -O2 for armv6 and armv7 is probably far more tested
>>>> than people generally expect. The share/mk/sys.mk change
>>>> goes back to -r319861 2017-Jun-12 . Previously the logic
>>>> would have caused -O use for armv6 or armv7 in MACHINE_ARCH .]
>>>
>>> r319861 doesn't look related here.
>>
>> # more /etc/make.conf 
>> .sinclude "/etc/make.nxb.conf"
>> USE_PACKAGE_DEPENDS=yes
>> BATCH=yes
>> WRKDIRPREFIX=/wrkdirs
>> PORTSDIR=/usr/ports
>> PACKAGES=/packages
>> DISTDIR=/distfiles
>> FORCE_PACKAGE=yes
>> PACKAGE_BUILDING=yes
>> PACKAGE_BUILDING_FLAVORS=yes
>> MACHINE=arm
>> MACHINE_ARCH=armv7
>> ARCH=${MACHINE_ARCH}
>> #### /usr/local/etc/poudriere.d/make.conf ####
>> #### /usr/ports/Mk/Scripts/ports_env.sh ####
>> _CCVERSION_9d218390=FreeBSD clang version 6.0.1 (tags/RELEASE_601/final 335540) (based on LLVM 6.0.1) Target: armv7-unknown-freebsd13.0-gnueabihf Thread model: posix InstalledDir: /nxb-bin/usr/bin
>> _ALTCCVERSION_9d218390=none
>> _CXXINTERNAL_9c45a5b1=FreeBSD clang version 6.0.1 (tags/RELEASE_601/final 335540) (based on LLVM 6.0.1) Target: armv7-unknown-freebsd13.0-gnueabihf Thread model: posix InstalledDir: /nxb-bin/usr/bin "/nxb-bin/usr/bin/ld" "--eh-frame-hdr" "-dynamic-linker" "/libexec/ld-elf.so.1" "--hash-style=both" "--enable-new-dtags" "-o" "a.out" "/usr/lib/crt1.o" "/usr/lib/crti.o" "/usr/lib/crtbegin.o" "-L/usr/lib" "/dev/null" "-lc++" "-lm" "-lgcc" "--as-needed" "-lgcc_s" "--no-as-needed" "-lc" "-lgcc" "--as-needed" "-lgcc_s" "--no-as-needed" "/usr/lib/crtend.o" "/usr/lib/crtn.o"
>> CC_OUTPUT_9d218390_58173849=yes
>> CC_OUTPUT_9d218390_9bdba57c=yes
>> CC_OUTPUT_9d218390_6a4fe7f5=yes
>> CC_OUTPUT_9d218390_6bcac02b=yes
>> CC_OUTPUT_9d218390_67d20829=yes
>> CC_OUTPUT_9d218390_bfa62e83=yes
>> CC_OUTPUT_9d218390_f0b4d593=yes
>> CC_OUTPUT_9d218390_308abb44=yes
>> CC_OUTPUT_9d218390_f00456e5=yes
>> CC_OUTPUT_9d218390_65ad290d=yes
>> CC_OUTPUT_9d218390_f2776b26=yes
>> CC_OUTPUT_9d218390_b2657cc3=yes
>> CC_OUTPUT_9d218390_380987f7=yes
>> CC_OUTPUT_9d218390_160933ec=yes
>> CC_OUTPUT_9d218390_fb62803b=yes
>> _OBJC_CCVERSION_9d218390=FreeBSD clang version 6.0.1 (tags/RELEASE_601/final 335540) (based on LLVM 6.0.1) Target: armv7-unknown-freebsd13.0-gnueabihf Thread model: posix InstalledDir: /nxb-bin/usr/bin
>> _OBJC_ALTCCVERSION_9d218390=none
>> ARCH=armv7
>> OPSYS=FreeBSD
>> _OSRELEASE=13.0-CURRENT
>> OSREL=13.0
>> OSVERSION=1300003
>> PYTHONBASE=/usr/local
>> _SMP_CPUS=32
>> CONFIGURE_MAX_CMD_LEN=262144
>> HAVE_PORTS_ENV=1
>> #### Misc Poudriere ####
>> GID=0
>> UID=0
>> PACKAGES=/packages
>>
>> # more /etc/src.conf
>> /etc/src.conf: No such file or directory
>>
>> # more Makefile
>> all:
>>        echo ${MACHINE_ARCH}
>>        echo ${MACHINE_CPUARCH}
>>        echo ${CFLAGS}
>>
>> # make
>> echo armv7
>> armv7
>> echo arm
>> arm
>> echo -O2 -pipe
>> -O2 -pipe
>>
>> # grep -r "\-O2" /usr/src/share/mk/
>> /usr/src/share/mk/sys.mk:CFLAGS		?=	-O2 -pipe
>>
>> # grep -r "\-pipe" /usr/src/share/mk/
>> /usr/src/share/mk/sys.mk:CFLAGS		?=	-O -pipe
>> /usr/src/share/mk/sys.mk:CFLAGS		?=	-O2 -pipe
>>
>> Those lines come from:
>>
>> .if ${MACHINE_CPUARCH} == "arm" || ${MACHINE_CPUARCH} == "mips"
>> CFLAGS          ?=      -O -pipe
>> .else
>> CFLAGS          ?=      -O2 -pipe
>> .endif
>>
>> So I used:
>>
>> # make -dA 2>2mmjnk.txt 1>1mmjnk.txt
>>
>> and looked for the first -pipe in 2mmjnk.txt. (It is also
>> the first -O , in this case -O2 .)
>>
>> . . .
>> Got 'C/mips(n32|64)?(el)?(hf)?/mips/:C/arm(v[67])?(eb)?/arm/:C/powerpc(64|spe)/powerpc/:C/riscv64(sf)?/riscv/' from '${__TO_CPUARCH}'}
>> Applying[MACHINE_ARCH] :C to "amd64"
>> Modifier pattern: "mips(n32|64)?(el)?(hf)?"
>> Modifier pattern: "mips"
>> Result[MACHINE_ARCH] of :C is "amd64"
>> Applying[MACHINE_ARCH] :C to "amd64"
>> Modifier pattern: "arm(v[67])?(eb)?"
>> Modifier pattern: "arm"
>> Result[MACHINE_ARCH] of :C is "amd64"
>> Applying[MACHINE_ARCH] :C to "amd64"
>> Modifier pattern: "powerpc(64|spe)"
>> Modifier pattern: "powerpc"
>> Result[MACHINE_ARCH] of :C is "amd64"
>> Applying[MACHINE_ARCH] :C to "amd64"
>> Modifier pattern: "riscv64(sf)?"
>> Modifier pattern: "riscv"
>> Result[MACHINE_ARCH] of :C is "amd64"
>> lhs = "amd64", rhs = "arm", op = ==
>> Got 'C/mips(n32|64)?(el)?(hf)?/mips/:C/arm(v[67])?(eb)?/arm/:C/powerpc(64|spe)/powerpc/:C/riscv64(sf)?/riscv/' from '${__TO_CPUARCH}'}
>> Applying[MACHINE_ARCH] :C to "amd64"
>> Modifier pattern: "mips(n32|64)?(el)?(hf)?"
>> Modifier pattern: "mips"
>> Result[MACHINE_ARCH] of :C is "amd64"
>> Applying[MACHINE_ARCH] :C to "amd64"
>> Modifier pattern: "arm(v[67])?(eb)?"
>> Modifier pattern: "arm"
>> Result[MACHINE_ARCH] of :C is "amd64"
>> Applying[MACHINE_ARCH] :C to "amd64"
>> Modifier pattern: "powerpc(64|spe)"
>> Modifier pattern: "powerpc"
>> Result[MACHINE_ARCH] of :C is "amd64"
>> Applying[MACHINE_ARCH] :C to "amd64"
>> Modifier pattern: "riscv64(sf)?"
>> Modifier pattern: "riscv"
>> Result[MACHINE_ARCH] of :C is "amd64"
>> lhs = "amd64", rhs = "mips", op = ==
>> ParseReadLine (172): 'CFLAGS            ?=      -O2 -pipe'
>> Global:CFLAGS = -O2 -pipe
>>
>> It looks like the poudriere builds based on /nxb-bin/. . .
>> See the amd64 MACHINE_ARCH in the context that shows:
>>
>> # uname -apKU
>> FreeBSD FBSDamd64TR 13.0-CURRENT FreeBSD 13.0-CURRENT arm armv7 1300003 1300003
> 
> 
> # grep "MACHINE_.*ARCH" 2mmjnk.txt
> ParseReadLine (17): 'MACHINE_CPUARCH=${MACHINE_ARCH:${__TO_CPUARCH}}'
> Global:MACHINE_CPUARCH = ${MACHINE_ARCH:${__TO_CPUARCH}}
> Applying[MACHINE_ARCH] :C to "amd64"
> Result[MACHINE_ARCH] of :C is "amd64"
> Applying[MACHINE_ARCH] :C to "amd64"
> Result[MACHINE_ARCH] of :C is "amd64"
> Applying[MACHINE_ARCH] :C to "amd64"
> Result[MACHINE_ARCH] of :C is "amd64"
> Applying[MACHINE_ARCH] :C to "amd64"
> Result[MACHINE_ARCH] of :C is "amd64"
> Applying[MACHINE_ARCH] :C to "amd64"
> Result[MACHINE_ARCH] of :C is "amd64"
> Applying[MACHINE_ARCH] :C to "amd64"
> Result[MACHINE_ARCH] of :C is "amd64"
> Applying[MACHINE_ARCH] :C to "amd64"
> Result[MACHINE_ARCH] of :C is "amd64"
> Applying[MACHINE_ARCH] :C to "amd64"
> Result[MACHINE_ARCH] of :C is "amd64"
> ParseReadLine (12): 'MACHINE_ARCH=armv7'
> Global:MACHINE_ARCH = armv7
> ParseReadLine (13): 'ARCH=${MACHINE_ARCH}'
> Global:ARCH = ${MACHINE_ARCH}
> ParseReadLine (2): '	echo ${MACHINE_ARCH}'
> ParseReadLine (3): '	echo ${MACHINE_CPUARCH}'
> MACHINE_CPUARCH  = ${MACHINE_ARCH:${__TO_CPUARCH}}
> MACHINE_ARCH     = armv7
> Applying[MACHINE_ARCH] :C to "armv7"
> Result[MACHINE_ARCH] of :C is "armv7"
> Applying[MACHINE_ARCH] :C to "armv7"
> Result[MACHINE_ARCH] of :C is "arm"
> Applying[MACHINE_ARCH] :C to "arm"
> Result[MACHINE_ARCH] of :C is "arm"
> Applying[MACHINE_ARCH] :C to "arm"
> Result[MACHINE_ARCH] of :C is "arm"
> MACHINE_CPUARCH  = ${MACHINE_ARCH:${__TO_CPUARCH}}
> MACHINE_ARCH     = armv7
> 
> Part of the issue is the ?= use from before the
> explicit MACHINE_ARCH assignment blocks any later
> ?= attempt form changing the result.
> 
> I'll note that any of the following executing before the
> explicitly MACHINE_ARCH assignment would apparently get
> the amd64 value from the /nxb-bin/. . . context, just like
> the .if conditional in under investigation:
> 
> # grep -r "^\..*MACHINE_ARCH" /usr/src/Makefile* /usr/src/share/mk/ /usr/ports/Mk/ | more
> /usr/src/Makefile.inc1:.if ${MACHINE_ARCH} != ${BUILD_ARCH}
> /usr/src/Makefile.inc1:.warning "To cross-build, set TARGET_ARCH ${MACHINE_ARCH} != ${BUILD_ARCH} ."
> /usr/src/Makefile.inc1:.if ${TARGET_ARCH} != ${MACHINE_ARCH} || ${TARGET} != ${MACHINE} || \
> /usr/src/Makefile.inc1:.if ${TARGET_ARCH} != ${MACHINE_ARCH} || ${BUILD_WITH_STRICT_TMPPATH} != 0
> /usr/src/Makefile.inc1:.elif ${TARGET_ARCH} != ${MACHINE_ARCH} && ${MK_ELFTOOLCHAIN_BOOTSTRAP} != "no"
> /usr/src/Makefile.inc1:.if ${TARGET} == ${MACHINE} && ${TARGET_ARCH} == ${MACHINE_ARCH}
> /usr/src/share/mk/bsd.arch.inc.mk:.elif exists(${.CURDIR}/Makefile.${MACHINE_ARCH})
> /usr/src/share/mk/bsd.arch.inc.mk:.include "Makefile.${MACHINE_ARCH}"
> /usr/src/share/mk/local.meta.sys.mk:.if empty(MACHINE_ARCH)
> /usr/src/share/mk/bsd.cpu.mk:. elif ${MACHINE_ARCH} == "sparc64"
> /usr/src/share/mk/bsd.cpu.mk:. elif ${MACHINE_ARCH} == "powerpc"
> /usr/src/share/mk/bsd.cpu.mk:. elif ${MACHINE_ARCH} == "powerpcspe"
> /usr/src/share/mk/bsd.cpu.mk:. elif ${MACHINE_ARCH} == "powerpc64"
> /usr/src/share/mk/bsd.cpu.mk:. elif ${MACHINE_ARCH} == "sparc64"
> /usr/src/share/mk/bsd.cpu.mk:. elif ${MACHINE_ARCH} == "powerpc"
> /usr/src/share/mk/bsd.cpu.mk:. elif ${MACHINE_ARCH} == "sparc64"
> /usr/src/share/mk/bsd.cpu.mk:. if ${MACHINE_ARCH:Mmips*el*} != ""
> /usr/src/share/mk/bsd.cpu.mk:. if ${MACHINE_ARCH:Mmips64*} != ""
> /usr/src/share/mk/bsd.cpu.mk:. elif ${MACHINE_ARCH:Mmipsn32*} != ""
> /usr/src/share/mk/bsd.cpu.mk:. if ${MACHINE_ARCH:Mmips*hf}
> /usr/src/share/mk/bsd.cpu.mk:. if ${MACHINE_ARCH:Marmv6*} != ""
> /usr/src/share/mk/bsd.cpu.mk:. if ${MACHINE_ARCH:Marmv7*} != ""
> /usr/src/share/mk/bsd.cpu.mk:. if ${MACHINE_ARCH:Marmv[67]*} == ""
> /usr/src/share/mk/bsd.cpu.mk:.if ${MACHINE_ARCH:Marmv[67]*} && defined(CPUTYPE) && ${CPUTYPE:M*soft*} != ""
> /usr/src/share/mk/bsd.cpu.mk:.if ${MACHINE_ARCH} == "powerpcspe"
> /usr/src/share/mk/bsd.cpu.mk:.if ${MACHINE_ARCH:Mriscv*sf}
> /usr/src/share/mk/bsd.endian.mk:.if ${MACHINE_ARCH} == "aarch64" || \
> /usr/src/share/mk/bsd.endian.mk:.elif ${MACHINE_ARCH} == "powerpc" || \
> /usr/ports/Mk/Uses/gnustep.mk:.if ${MACHINE_ARCH} == "i386"
> 
> (I have not tried to figure out which have a chance of
> executing before the MACHINE_ARCH explicit assignment.)
> 
> ===
> Mark Millard
> marklmi at yahoo.com
> ( dsl-only.net went
> away in early 2018-Mar)
> 

I think the real problem here is that Poudriere is setting MACHINE_ARCH
in make.conf and sys.mk loads make.conf *after* checking MACHINE_CPUARCH
(derived from MACHINE_ARCH) to determine CFLAGS; The .if is expanding
MACHINE_CPUARCH before make.conf is included.

We probably need a make-env.conf thing like src-env.conf to allow
modifying sys.mk earlier.


-- 
Regards,
Bryan Drewery
signature.asc (application/pgp-signature, 488 B)