#65668 closed defect (fixed)
OpenBLAS @0.3.21 has a problem on PPC Mac OS X 10.4.11, Tiger, with 'mktemp'
Reported by: | ballapete (Peter "Pete" Dyballa) | Owned by: | michaelld (Michael Dickens) |
---|---|---|---|
Priority: | Normal | Milestone: | |
Component: | ports | Version: | 2.7.2 |
Keywords: | tiger haspatch | Cc: | NicosPavlov |
Port: | OpenBLAS |
Description
gmake: Entering directory '/opt/local/var/macports/build/_opt_local_var_macports_sources_nue.de.rsync.macports.org_macports_release_tarballs_ports_math_OpenBLAS/OpenBLAS/work/xianyi-OpenBLAS-b89fb70' usage: mktemp [-d] [-q] [-t prefix] [-u] template ... mktemp [-d] [-q] [-u] -t prefix ./c_check: line 248: /a.c: Permission denied usage: mktemp [-d] [-q] [-t prefix] [-u] template ... mktemp [-d] [-q] [-u] -t prefix ./c_check: line 248: /a.c: Permission denied usage: mktemp [-d] [-q] [-t prefix] [-u] template ... mktemp [-d] [-q] [-u] -t prefix ./c_check: line 248: /a.c: Permission denied gmake[1]: Entering directory '/opt/local/var/macports/build/_opt_local_var_macports_sources_nue.de.rsync.macports.org_macports_release_tarballs_ports_math_OpenBLAS/OpenBLAS/work/xianyi-OpenBLAS-b89fb70/interface'
The code is:
244 case "$data" in 245 *HAVE_C11*) 246 tmpd=`mktemp -d` 247 tmpf="$tmpd/a.c" 248 printf "#include <stdatomic.h>\nint main(void){}\n" >> "$tmpf" 249 args=" -c -o $tmpf.o $tmpf" 250 c11_atomics=1 251 { 252 $compiler_name $flags $args >/dev/null 2>&1 253 } || { 254 c11_atomics=0 255 } 256 257 rm -rf "$tmpd" 258 ;; 259 esac
My PowerBook G4
has:
-rwxr-xr-x 1 root admin 13716 19. Jan 2022 /opt/local/lib/bash/mktemp lrwxr-xr-x 1 root admin 22 7. Jul 12:06 /opt/local/libexec/gnubin/mktemp -> /opt/local/bin/gmktemp -r-xr-xr-x 1 root wheel 14076 21. Mär 2005 /usr/bin/mktemp
Obviously Apples
/usr/bin/mktemp` is being used:
pete 241 /\ /usr/bin/mktemp -d usage: mktemp [-d] [-q] [-t prefix] [-u] template ... mktemp [-d] [-q] [-u] -t prefix Exit 1 pete 242 /\ /opt/local/libexec/gnubin/mktemp -d /tmp/tmp.vqK25Du8tm pete 243 /\ rmdir /tmp/tmp.vqK25Du8tm
Using /usr/bin/mktemp
it would need an additional -d /tmp/tmpdir${$}
or such argument.
Later this warning appears:
gmake[1]: Entering directory '/opt/local/var/macports/build/_opt_local_var_macports_sources_nue.de.rsync.macports.org_macports_release_tarballs_ports_math_OpenBLAS/OpenBLAS/work/xianyi-OpenBLAS-b89fb70/driver/others' /opt/local/var/macports/build/_opt_local_var_macports_sources_nue.de.rsync.macports.org_macports_release_tarballs_ports_math_OpenBLAS/OpenBLAS/work/compwrap/cc/opt/local/bin/gcc-mp-7 -Os -arch ppc -O3 -DMAX_STACK_ALLOC=2048 -Wall -m32 -DF_INTERFACE_GFORT -fPIC -DNO_AVX -DNO_AVX2 -DNO_AVX512 -DMAX_CPU_NUMBER=1 -DMAX_PARALLEL_NUMBER=1 -DBUILD_SINGLE=1 -DBUILD_DOUBLE=1 -DBUILD_COMPLEX=1 -DBUILD_COMPLEX16=1 -DVERSION=\"1\" -force_cpusubtype_ALL -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME -DASMNAME=_ -DASMFNAME=__ -DNAME=_ -DCNAME= -DCHAR_NAME=\"_\" -DCHAR_CNAME=\"\" -DNO_AFFINITY -I. -O3 -DSMALL_MATRIX_OPT -DMAX_STACK_ALLOC=2048 -Wall -m32 -DF_INTERFACE_GFORT -fPIC -DNO_AVX -DNO_AVX2 -DNO_AVX512 -DMAX_CPU_NUMBER=1 -DMAX_PARALLEL_NUMBER=1 -DBUILD_BFLOAT16 -DBUILD_SINGLE=1 -DBUILD_DOUBLE=1 -DBUILD_COMPLEX=1 -DBUILD_COMPLEX16=1 -DVERSION=\"1\" -force_cpusubtype_ALL -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME -DASMNAME=_memory -DASMFNAME=_memory_ -DNAME=memory_ -DCNAME=memory -DCHAR_NAME=\"memory_\" -DCHAR_CNAME=\"memory\" -DNO_AFFINITY -I../.. -c memory.c -o memory.o memory.c:106:2: warning: #warning BUFFER_SIZE is too small for P, Q, and R of SGEMM - large calculations may crash ! [-Wcpp] #warning BUFFER_SIZE is too small for P, Q, and R of SGEMM - large calculations may crash ! ^~~~~~~ memory.c:111:2: warning: #warning BUFFER_SIZE is too small for P, Q, and R of DGEMM - large calculations may crash ! [-Wcpp] #warning BUFFER_SIZE is too small for P, Q, and R of DGEMM - large calculations may crash ! ^~~~~~~ memory.c:116:2: warning: #warning BUFFER_SIZE is too small for P, Q, and R of CGEMM - large calculations may crash ! [-Wcpp] #warning BUFFER_SIZE is too small for P, Q, and R of CGEMM - large calculations may crash ! ^~~~~~~ memory.c:121:2: warning: #warning BUFFER_SIZE is too small for P, Q, and R of ZGEMM - large calculations may crash ! [-Wcpp] #warning BUFFER_SIZE is too small for P, Q, and R of ZGEMM - large calculations may crash ! ^~~~~~~ /opt/local/var/macports/build/_opt_local_var_macports_sources_nue.de.rsync.macports.org_macports_release_tarballs_ports_math_OpenBLAS/OpenBLAS/work/compwrap/cc/opt/local/bin/gcc-mp-7 -Os -arch ppc -O3 -DMAX_STACK_ALLOC=2048 -Wall -m32 -DF_INTERFACE_GFORT -fPIC -DNO_AVX -DNO_AVX2 -DNO_AVX512 -DMAX_CPU_NUMBER=1 -DMAX_PARALLEL_NUMBER=1 -DBUILD_SINGLE=1 -DBUILD_DOUBLE=1 -DBUILD_COMPLEX=1 -DBUILD_COMPLEX16=1 -DVERSION=\"1\" -force_cpusubtype_ALL -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME -DASMNAME=_ -DASMFNAME=__ -DNAME=_ -DCNAME= -DCHAR_NAME=\"_\" -DCHAR_CNAME=\"\" -DNO_AFFINITY -I. -O3 -DSMALL_MATRIX_OPT -DMAX_STACK_ALLOC=2048 -Wall -m32 -DF_INTERFACE_GFORT -fPIC -DNO_AVX -DNO_AVX2 -DNO_AVX512 -DMAX_CPU_NUMBER=1 -DMAX_PARALLEL_NUMBER=1 -DBUILD_BFLOAT16 -DBUILD_SINGLE=1 -DBUILD_DOUBLE=1 -DBUILD_COMPLEX=1 -DBUILD_COMPLEX16=1 -DVERSION=\"1\" -force_cpusubtype_ALL -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME -DASMNAME=_xerbla -DASMFNAME=_xerbla_ -DNAME=xerbla_ -DCNAME=xerbla -DCHAR_NAME=\"xerbla_\" -DCHAR_CNAME=\"xerbla\" -DNO_AFFINITY -I../.. -c xerbla.c -o xerbla.o
There are more warnings, about variables used uninitialised or that could change in value when inherently converted between different types, or were left unused or ignored. (Compilation still going on.)
Attachments (1)
Change History (11)
comment:1 Changed 2 years ago by ballapete (Peter "Pete" Dyballa)
comment:2 follow-up: 3 Changed 2 years ago by kencu (Ken)
the usual drill for this is to add a dep on coreutils and patch in gmktemp
comment:3 Changed 2 years ago by ballapete (Peter "Pete" Dyballa)
Replying to kencu:
On PPC Leopard, Mac OS X 10.5.8
I changed mktemp
to gmktemp
and the build suceeded. I'll attach my simple patch afterwards.
Changed 2 years ago by ballapete (Peter "Pete" Dyballa)
Attachment: | patch-MacOSX-mktemp.diff added |
---|
Simple patch to change too old mktemp to gmktemp in c_check shell script
comment:4 Changed 2 years ago by mascguy (Christopher Nielsen)
Cc: | NicosPavlov added; nicos@… michaelld@… removed |
---|---|
Keywords: | haspatch added |
Owner: | set to michaelld |
Status: | new → assigned |
comment:5 Changed 2 years ago by NicosPavlov <pavillon.nicolas@…>
comment:6 Changed 2 years ago by NicosPavlov
Thank you for the patch, I left the ticket open in order to confirm that it indeed fixes the issue before closing it.
comment:8 Changed 2 years ago by ballapete (Peter "Pete" Dyballa)
OpenBLAS built nevertheless overnight.
comment:9 Changed 2 years ago by NicosPavlov <pavillon.nicolas@…>
comment:10 Changed 2 years ago by NicosPavlov
Resolution: | → fixed |
---|---|
Status: | assigned → closed |
Thanks for the comment, I added the dependency. I'll mark the ticket as fixed at this point.
The
mktemp
problems repeats near the end: