Bug 199929
| Summary: | [NEW PORT] math/cryptominisat | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | Ports & Packages | Reporter: | Gleb Popov <arrowd> | ||||||
| Component: | Individual Port(s) | Assignee: | Jan Beich <jbeich> | ||||||
| Status: | Closed FIXED | ||||||||
| Severity: | Affects Only Me | CC: | jbeich | ||||||
| Priority: | --- | ||||||||
| Version: | Latest | ||||||||
| Hardware: | Any | ||||||||
| OS: | Any | ||||||||
| Attachments: |
|
||||||||
Comment on attachment 156338 [details] Patch with new port $ portlint -AC WARN: /usr/ports/math/cryptominisat/pkg-plist: [8]: installing shared libraries, please define USE_LDCONFIG as appropriate FATAL: /usr/ports/math/cryptominisat/pkg-plist: the last line has to be terminated by \n. >+++ math/cryptominisat/Makefile 2015-05-02 22:20:57.733228258 +0300 [...] >+.include <bsd.port.options.mk> >+ >+.if ${PORT_OPTIONS:MPYTHON} New ports should use option helpers if possible. https://www.freebsd.org/doc/en/books/porters-handbook/makefile-options.html#options-helpers >+USES+= python:build USES=python with no version specified is for ports that support both python2 and python3. CMakeLists.txt seems to disagree: find_package (PythonInterp 2.7) find_package (PythonLibs 2.7) >+PLIST_SUB+= EGGFILE=pycryptosat-${PORTVERSION}-py${PYTHON_VER}.egg-info >+.else >+CMAKE_ARGS+= -DPYTHON_EXECUTABLE=0 >+.endif >++++ python/CMakeLists.txt >+@@ -12,7 +12,7 @@ add_custom_command(OUTPUT ${OUTPUT}/time >+ >+ add_custom_target(pytarget ALL DEPENDS ${OUTPUT}/timestamp) >+ >+-install(CODE "execute_process(COMMAND ${PYTHON_EXECUTABLE} ${SETUP_PY} install --record files.txt)") >++install(CODE "execute_process(COMMAND ${PYTHON_EXECUTABLE} ${SETUP_PY} install -c -O1 --prefix=${CMAKE_INSTALL_PREFIX} --root=${FREEBSD_STAGEDIR} --record files.txt)") Why not create a separate USE_PYTHON=distutils port that depends on this one? >+++ math/cryptominisat/pkg-plist 2015-05-02 16:26:25.333693829 +0300 >@@ -0,0 +1,10 @@ [...] >+lib/libcryptominisat4.so Maybe set SOVERSION, so the library path can be cached by ldconfig(8) and used not only via RPATH. Also fails to build on FreeBSD 8.4R and 9.3R: cc1plus: error: unrecognized command line option "-std=c++11" (In reply to Jan Beich from comment #1) >WARN: /usr/ports/math/cryptominisat/pkg-plist: [8]: installing shared libraries, please define USE_LDCONFIG as appropriate >Maybe set SOVERSION, so the library path can be cached by ldconfig(8) and used not only via RPATH. Yup, but what's the proper way to do this? Make a symlink at post-install? >USES=python with no version specified is for ports that support both python2 and python3. CMakeLists.txt seems to disagree: As with find_package(Boost 1.46) case, this version number is a minimum version required. Since cryptominisat author doesn't check for actual version of python that have been found, i thought it should work with any python that is 2.7+. > Why not create a separate USE_PYTHON=distutils port that depends on this one? I don't see why it's required. The setup.py script is still configured by CMake, so, from my understanding, i would have to run either cmake or setup.py manually. Regarding other problems, i'll fix them and resubmit the port. Thanks for review. Created attachment 156406 [details] PYTHON as separate ports, v0 (In reply to 6yearold from comment #3) > (In reply to Jan Beich from comment #1) >>WARN: /usr/ports/math/cryptominisat/pkg-plist: [8]: installing shared >>libraries, please define USE_LDCONFIG as appropriate >>Maybe set SOVERSION, so the library path can be cached by ldconfig(8) >>and used not only via RPATH. > Yup, but what's the proper way to do this? Make a symlink at post-install? After testing it works fine with just USE_LDCONFIG. Broken ldconfig -r output can be ignored as it doesn't affect LIB_DEPENDS implementation for a few years. If you still want SOVERSION try: set_target_properties(libcryptominisat4 PROPERTIES ... SOVERSION 0 ) >>USES=python with no version specified is for ports that support both >>python2 and python3. CMakeLists.txt seems to disagree: > > As with find_package(Boost 1.46) case, this version number is a minimum version > required. Since cryptominisat author doesn't check for actual version of python > that have been found, i thought it should work with any python that is 2.7+. Try building with: # in make.conf DEFAULT_VERSIONS = python=3.4 which would either not find suitable python version or fail as pycryptosat.cpp:68:10: error: use of undeclared identifier 'PyInt_Check' if (!IS_INT(lit)) { ^ pycryptosat.cpp:21:21: note: expanded from macro 'IS_INT' #define IS_INT(x) (PyInt_Check(x) || PyLong_Check(x)) ^ pycryptosat.cpp:366:11: error: no member named 'ob_type' in 'Solver' self->ob_type->tp_free((PyObject*)self); ~~~~ ^ pycryptosat.cpp:409:5: error: cannot initialize a member subobject of type 'Py_ssize_t' (aka 'long') with an lvalue of type 'const char [19]' "pycryptosat.Solver", /*tp_name*/ ^~~~~~~~~~~~~~~~~~~~ pycryptosat.cpp:412:5: error: cannot initialize a member subobject of type 'printfunc' (aka 'int (*)(PyObject *, FILE *, int)') with an rvalue of type 'destructor' (aka 'void (*)(PyObject *)'): different number of parameters (3 vs 1) (destructor)Solver_dealloc, /*tp_dealloc*/ ^~~~~~~~~~~~~~~~~~~~~~~~~~ pycryptosat.cpp:427:5: error: cannot initialize a member subobject of type 'const char *' with an rvalue of type 'unsigned long' Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE, /*tp_flags*/ ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ /usr/local/include/python3.4/object.h:647:29: note: expanded from macro 'Py_TPFLAGS_DEFAULT' #define Py_TPFLAGS_DEFAULT ( \ ^ pycryptosat.cpp:428:5: error: cannot initialize a member subobject of type 'traverseproc' (aka 'int (*)(PyObject *, visitproc, void *)') with an lvalue of type 'const char [15]' "Solver objects", /* tp_doc */ ^~~~~~~~~~~~~~~~ pycryptosat.cpp:435:5: error: cannot initialize a member subobject of type 'struct PyMemberDef *' with an lvalue of type 'PyMethodDef [4]' Solver_methods, /* tp_methods */ ^~~~~~~~~~~~~~ pycryptosat.cpp:436:5: error: cannot initialize a member subobject of type 'struct PyGetSetDef *' with an lvalue of type 'PyMemberDef [1]' Solver_members, /* tp_members */ ^~~~~~~~~~~~~~ pycryptosat.cpp:443:5: error: cannot initialize a member subobject of type 'allocfunc' (aka 'PyObject *(*)(struct _typeobject *, Py_ssize_t)') with an rvalue of type 'initproc' (aka 'int (*)(PyObject *, PyObject *, PyObject *)'): different number of parameters (2 vs 3) (initproc)Solver_init, /* tp_init */ ^~~~~~~~~~~~~~~~~~~~~ pycryptosat.cpp:445:5: error: cannot initialize a member subobject of type 'freefunc' (aka 'void (*)(void *)') with an lvalue of type 'PyObject *(PyTypeObject *, PyObject *, PyObject *)': different number of parameters (1 vs 3) Solver_new, /* tp_new */ ^~~~~~~~~~ pycryptosat.cpp:458:9: error: non-void function 'initpycryptosat' should return a value [-Wreturn-type] return; ^ pycryptosat.cpp:460:9: error: use of undeclared identifier 'Py_InitModule3' m = Py_InitModule3("pycryptosat", module_methods, ^ Also, USES+=python:build doesn't account for libpython dependency. $ readelf -d .../pycryptosat.so | fgrep NEEDED 0x0000000000000001 (NEEDED) Shared library: [libcryptominisat4.so] 0x0000000000000001 (NEEDED) Shared library: [libpython2.7.so.1] 0x0000000000000001 (NEEDED) Shared library: [libc++.so.1] 0x0000000000000001 (NEEDED) Shared library: [libcxxrt.so.1] 0x0000000000000001 (NEEDED) Shared library: [libm.so.5] 0x0000000000000001 (NEEDED) Shared library: [libgcc_s.so.1] 0x0000000000000001 (NEEDED) Shared library: [libc.so.7] >> Why not create a separate USE_PYTHON=distutils port that depends on this one? > I don't see why it's required. The setup.py script is still configured by > CMake, so, from my understanding, i would have to run either cmake or setup.py > manually. It was merely a suggestion (not requirement) to: - let the ports framework set the build environment for you - provide python bindings for package users I'm attaching an example with all of the mentioned issues addressed. Hopefully, slave deduplication doesn't make the approach hard to understand. (In reply to Jan Beich from comment #4) >After testing it works fine with just USE_LDCONFIG. Broken ldconfig -r output can be ignored as it doesn't affect LIB_DEPENDS implementation for a few years. Linking with libcryptominisat4.so requires SOVERSION too, so it's better to keep it. >I'm attaching an example with all of the mentioned issues addressed. Ok, thanks. Since you've done all work for me, then you could commit it? A commit references this bug: Author: jbeich Date: Tue May 5 21:54:33 UTC 2015 New revision: 385495 URL: https://svnweb.freebsd.org/changeset/ports/385495 Log: math/cryptominisat, math/py-cryptominisat: add new ports CryptoMiniSat is a modern, multi-threaded, feature-rich, simplifying SAT solver, featuring over 100 configurable parameters to tune to specific need, collection of statistical data to MySQL database + javascript-based visualization of it and clean C++ and python interfaces. WWW: http://www.msoos.org/cryptominisat4/ PR: 199929 Submitted by: 6yearold@gmail.com (based on) Changes: head/math/Makefile head/math/cryptominisat/ head/math/cryptominisat/Makefile head/math/cryptominisat/distinfo head/math/cryptominisat/files/ head/math/cryptominisat/files/patch-cryptominisat4_CMakeLists.txt head/math/cryptominisat/files/patch-cryptominisat4_bva.cpp head/math/cryptominisat/pkg-descr head/math/cryptominisat/pkg-plist head/math/py-cryptominisat/ head/math/py-cryptominisat/Makefile head/math/py-cryptominisat/files/ head/math/py-cryptominisat/files/patch-pycryptosat.cpp Thanks. Committed. Hopefully the port makes it into the next periodic package build. |
Created attachment 156338 [details] Patch with new port CryptoMiniSat is a modern, multi-threaded, feature-rich, simplifying SAT solver, featuring over 100 configurable parameters to tune to specific need, collection of statistical data to MySQL database + javascript-based visualization of it and clean C++ and python interfaces.