Re: ICC 15

[Thomas Orgis <thomas.orgis%uni-hamburg.de@localhost>]

Am Fri, 2 Oct 2015 09:51:44 -0500
schrieb Jason Bacon <jwbacon%tds.net@localhost>: 

> 
> FYI, I was able to successfully bootstrap using ICC 15 on CentOS 6.7.  

I started to try to get some pkgsrc running with ICC 15 some time ago.
I'm not sure right now if there were bootstrap problems, but ...

> icc: command line remark #10148: option '-static-libcxa' not supported  

I felt I needed to remove -static-libcxa from
mk/wrapper/cmd-sink-icc-cc to be able to get perl built. Something with
non-PIC binaries entering where they are not desired.

My current task is to get devel/netcdf installed. So far editors/nano,
devel/hdf, devel/hdf5, devel/hdf5-c++ worked, along with some
dependency stuff. Oh, yes, dependency stuff is what prevents
devel/netcdf so far:

checking if icc supports -c -o file.o... (cached) yes
checking whether the icc linker (/scratch/sw/work/intel-15.0.3_impi-5.0.3/pkgsrc-2015Q2/www/curl/work/.wrapper/bin/ld -m elf_x86_64) supports shared libraries... yes
checking whether -lc should be explicitly linked in... /bin/rm: cannot remove 'conftest*': No such file or directory
no
checking dynamic linker characteristics... GNU/Linux ld.so
checking how to hardcode library paths into programs... immediate
checking whether stripping libraries is possible... yes
checking if libtool supports shared libraries... yes
checking whether to build shared libraries... yes
checking whether to build static libraries... yes
checking whether to build shared libraries with -version-info... yes
checking whether to build shared libraries with -no-undefined... no
checking whether to build shared libraries with -mimpure-text... no
checking whether to build shared libraries with PIC... yes
checking whether to build static libraries with PIC... yes
checking whether to build shared libraries only... no
checking whether to build static libraries only... no
checking for inline... inline
checking if compiler is DEC/Compaq/HP C... no
checking if compiler is HP-UX C... no
checking if compiler is IBM C... no
checking if compiler is Intel C... yes
checking if compiler is clang... no
checking if compiler is GNU C... no
checking if compiler is LCC... no
checking if compiler is SGI MIPSpro C... no
checking if compiler is SGI MIPS C... no
checking if compiler is SunPro C... no
checking if compiler is Tiny C... no
checking if compiler is Watcom C... no
checking if compiler accepts debug disabling options... yes
configure: compiler options added: 
checking if compiler optimizer assumed setting might be used... yes
checking if compiler accepts optimizer enabling options... yes
configure: compiler options added: -O2
checking if compiler accepts strict warning options... yes
configure: compiler options added: -fno-omit-frame-pointer -fno-strict-aliasing -fp-model precise -vec-report0 
checking if compiler halts on compilation errors... yes
checking if compiler halts on negative sized arrays... yes
checking if compiler halts on function prototype mismatch... no
configure: error: compiler does not halt on function prototype mismatch.
*** Error code 1

Stop.
bmake[2]: stopped in /scratch/sw/work/intel-15.0.3_impi-5.0.3/pkgsrc-tree-2015Q2/pkgsrc/www/curl
*** Error code 1

Stop.
bmake[1]: stopped in /scratch/sw/work/intel-15.0.3_impi-5.0.3/pkgsrc-tree-2015Q2/pkgsrc/www/curl
*** Error code 1

It fails with www/curl. Maybe I can fix that up tonight. I really just
wanted to test some software that needs NetCDF …

… ah, yes: I need to change www/curl/patches/patch-aa:

diff -ru pkgsrc-2015Q2/www/curl/distinfo pkgsrc/www/curl/distinfo
--- pkgsrc-2015Q2/www/curl/distinfo	2015-06-03 14:00:06.000000000 +0200
+++ pkgsrc/www/curl/distinfo	2015-10-03 20:25:26.083755229 +0200
@@ -3,7 +3,7 @@
 SHA1 (curl-7.42.1.tar.bz2) = f65708915875b8cb35edb51d8dd31440dc02fbd3
 RMD160 (curl-7.42.1.tar.bz2) = 76d5b23fae60356342e2bac2e4c706ed544d4adf
 Size (curl-7.42.1.tar.bz2) = 3327304 bytes
-SHA1 (patch-aa) = 59ec0be3ac90470fdc5935881da6a14dbab9d378
+SHA1 (patch-aa) = 2c46dc013748a2deb0b1bc79d14526c1913870d3
 SHA1 (patch-curl-config.in) = fd87c97b601a6b9269f67fbc066604ee7e22570e
 SHA1 (patch-lib_hostcheck.c) = 9faf94f44703c7d37377fd3af319ca5c27df34c2
 SHA1 (patch-lib_http2.c) = 4ba0164ffdba714c620daccbf80eedd51562acf4
diff -ru pkgsrc-2015Q2/www/curl/patches/patch-aa pkgsrc/www/curl/patches/patch-aa
--- pkgsrc-2015Q2/www/curl/patches/patch-aa	2015-04-22 16:35:21.000000000 +0200
+++ pkgsrc/www/curl/patches/patch-aa	2015-10-03 20:15:04.377468388 +0200
@@ -18,7 +18,7 @@
  
        #
 -  if test "$compiler_id" != "unknown"; then
-+  if false; then
++  if test "$compiler_id" != "unknown"; then # ThOr: re-enabled for ICC
      #
      if test "$compiler_id" = "GNU_C" ||
        test "$compiler_id" = "CLANG"; then

The patch claimed to fix something related to krb5, but also contained
that disabling part of curls's compiler machinery, resulting apparently
in _not_ adding -std=gnu89, which sabotages the compiler check later.


alrighty then,

Thomas

-- 
Dr. Thomas Orgis
Universität Hamburg
RRZ / Zentrale Dienste / HPC
Schlüterstr. 70
20146 Hamburg
Tel.: 040/42838 8826
Fax: 040/428 38 6270