[SCIP_Papilo] up scip papilo to 9.2.1

[matbesancon]

No description provided.

[giordano]

This package is hitting ICEs with any version of GCC.

[imciner2]

[22:25:56] cd /workspace/srcdir/scipoptsuite-9.2.1/build/papilo/test && /opt/bin/aarch64-linux-gnu-libgfortran4-cxx03/aarch64-linux-gnu-g++ --sysroot=/opt/aarch
64-linux-gnu/aarch64-linux-gnu/sys-root/ -DBOOST_IOSTREAMS_DYN_LINK -DBOOST_IOSTREAMS_NO_LIB -DBOOST_PROGRAM_OPTIONS_DYN_LINK -DBOOST_PROGRAM_OPTIONS_NO_LIB -DB
OOST_SERIALIZATION_DYN_LINK -DBOOST_SERIALIZATION_NO_LIB -I/workspace/srcdir/scipoptsuite-9.2.1/soplex/src -I/workspace/srcdir/scipoptsuite-9.2.1/scip/src -I/wo
rkspace/srcdir/scipoptsuite-9.2.1/zimpl/src -I/workspace/srcdir/scipoptsuite-9.2.1/papilo/src -I/workspace/srcdir/scipoptsuite-9.2.1/ug/src -I/workspace/srcdir/
scipoptsuite-9.2.1/gcg/src -I/workspace/srcdir/scipoptsuite-9.2.1/build/papilo -DTHREADLOCAL='' -Wno-psabi -O3 -DNDEBUG -fPIE -fvisibility=hidden -fvisibility-i
nlines-hidden -ffp-contract=off -Wno-shadow -Wall -std=gnu++14 -MD -MT papilo/test/CMakeFiles/unit_test.dir/TestMain.cpp.o -MF CMakeFiles/unit_test.dir/TestMain
.cpp.o.d -o CMakeFiles/unit_test.dir/TestMain.cpp.o -c /workspace/srcdir/scipoptsuite-9.2.1/papilo/test/TestMain.cpp
[22:26:04] In file included from /workspace/srcdir/scipoptsuite-9.2.1/papilo/test/TestMain.cpp:25:0:
[22:26:04] /workspace/srcdir/scipoptsuite-9.2.1/papilo/src/papilo/external/catch/catch_amalgamated.hpp: In function ‘uint64_t Catch::ulpDistance(FP, FP) [with F
P = float]’:
[22:26:04] /workspace/srcdir/scipoptsuite-9.2.1/papilo/src/papilo/external/catch/catch_amalgamated.hpp:9705:5: internal compiler error: in aarch64_select_cc_mod
e, at config/aarch64/aarch64.c:4676
[22:26:04]      }
[22:26:04]      ^
[22:26:04] 0xd99a10 aarch64_select_cc_mode(rtx_code, rtx_def*, rtx_def*)
[22:26:04] 	/workspace/srcdir/gcc-7.1.0/gcc/config/aarch64/aarch64.c:4676
[22:26:04] 0xfa9cf0 simplify_set
[22:26:04] 	/workspace/srcdir/gcc-7.1.0/gcc/combine.c:6696
[22:26:04] 0xfa9cf0 combine_simplify_rtx
[22:26:04] 	/workspace/srcdir/gcc-7.1.0/gcc/combine.c:6217
[22:26:04] 0xfab728 subst
[22:26:04] 	/workspace/srcdir/gcc-7.1.0/gcc/combine.c:5517
[22:26:04] 0xfad2c8 try_combine
[22:26:04] 	/workspace/srcdir/gcc-7.1.0/gcc/combine.c:3332
[22:26:04] 0xfb2f89 combine_instructions
[22:26:04] 	/workspace/srcdir/gcc-7.1.0/gcc/combine.c:1372
[22:26:04] 0xfb2f89 rest_of_handle_combine
[22:26:04] 	/workspace/srcdir/gcc-7.1.0/gcc/combine.c:14653
[22:26:04] 0xfb2f89 execute
[22:26:04] 	/workspace/srcdir/gcc-7.1.0/gcc/combine.c:14698
[22:26:04] Please submit a full bug report,
[22:26:04] with preprocessed source if appropriate.
[22:26:04] Please include the complete backtrace with any bug report.
[22:26:04] See <https://gcc.gnu.org/bugs/> for instructions.
[22:26:04] make[2]: *** [papilo/test/CMakeFiles/unit_test.dir/build.make:76: papilo/test/CMakeFiles/unit_test.dir/TestMain.cpp.o] Error 1
[22:26:04] make[2]: Leaving directory '/workspace/srcdir/scipoptsuite-9.2.1/build'

This ICE looks like it is in the test directory, perhaps we can just disable building the tests and see if that lets the builds complete? Looking at their CMake, I think that is just done by adding -DBUILD_TESTING=OFF to the CMake command.

[imciner2]

And not related to the problem at hand, but I see there is an option for a HiGHS interface in papilo, would it be worth turning that on now?

[matbesancon]

would it be worth turning that on now?

Yes I think it will be worth creating a separate build for this

[imciner2]

Yes I think it will be worth creating a separate build for this

Oh, it has to be separate from the normal build and isn't just an option the user can choose at runtime?

[matbesancon]

Indeed, SCIP works with an abstract linear programming interface, only one corresponding to the current LP solver is compiled. Otherwise this would also bloat the binary size

[imciner2]

Indeed, SCIP works with an abstract linear programming interface, only one corresponding to the current LP solver is compiled. Otherwise this would also bloat the binary size

Makes sense. If you do make a new build for this, it would be good to do it with the common build script structure so that they can share as much of the code as possible.

[matbesancon]

yes for sure, for now I went the lazy version with SCIP{_PaPILO}, it will be good to put things in common