Re: [Rdkit-devel] seg faults running tests on a mac
Open-Source Cheminformatics and Machine Learning
Brought to you by:
glandrum
|
From: Cyrus H. <cyr...@gm...> - 2015-05-16 14:20:08
|
Following up on my own message... I noticed that the python executable and python libraries found by cmake don't match. cmake finds the homebrew installed python binary, but the system installed libraries. If I remove the homebrew python I can at least get the tests to pass (but run into problems with various other missing python packages). A step in the right direction. On Sat, May 16, 2015 at 10:11 AM, Cyrus Harmon <cyr...@gm...> wrote: > Hi Greg, > > Sorry it's been a while since I first reported my troubles on a Mac. I > continue to have problems on my MacBook Pro, but things seem to work fine > on linux for me. > > The problem is that many of the tests fail with a segfault and there > doesn't seem to much information in any of the log files about the problem. > In an effort to figure out what's going on here, I've been trying to run > things under gdb. > > If I run the testBV.py test under gdb I see the following: > > Program received signal EXC_BAD_ACCESS, Could not access memory. > Reason: KERN_INVALID_ADDRESS at address: 0x0000000000000000 > 0x00000001029e7cb9 in PyInt_FromLong () > (gdb) bt > #0 0x00000001029e7cb9 in PyInt_FromLong () > #1 0x0000000102844908 in > boost::python::detail::keywords<1ul>::operator=<int> () > #2 0x000000010284277a in init_module_rdBase () > #3 0x0000000102b2f391 in boost::python::handle_exception_impl () > #4 0x0000000102b303b9 in boost::python::detail::init_module () > #5 0x00000001000a04ea in _PyImport_LoadDynamicModule () > #6 0x00000001000a01d2 in import_submodule () > #7 0x000000010009fd9c in load_next () > #8 0x000000010009df99 in PyImport_ImportModuleLevel () > #9 0x0000000100083459 in builtin___import__ () > #10 0x000000010000e3b6 in PyObject_Call () > #11 0x000000010008dd9b in PyEval_CallObjectWithKeywords () > #12 0x000000010008a0b7 in PyEval_EvalFrameEx () > #13 0x0000000100087df9 in PyEval_EvalCodeEx () > #14 0x000000010008786f in PyEval_EvalCode () > #15 0x000000010009ca46 in PyImport_ExecCodeModuleEx () > #16 0x000000010009f731 in load_source_module () > #17 0x000000010009f9f2 in load_package () > #18 0x00000001000a01d2 in import_submodule () > #19 0x000000010009fd9c in load_next () > #20 0x000000010009df99 in PyImport_ImportModuleLevel () > #21 0x0000000100083459 in builtin___import__ () > #22 0x000000010000e3b6 in PyObject_Call () > #23 0x000000010008dd9b in PyEval_CallObjectWithKeywords () > #24 0x000000010008a0b7 in PyEval_EvalFrameEx () > #25 0x0000000100087df9 in PyEval_EvalCodeEx () > #26 0x000000010008786f in PyEval_EvalCode () > #27 0x00000001000a7a2a in run_mod () > #28 0x00000001000a7acd in PyRun_FileExFlags () > #29 0x00000001000a766a in PyRun_SimpleFileExFlags () > #30 0x00000001000b8d6f in Py_Main () > #31 0x00007fff947ec5c9 in start () > > So it looks to me like things are sort of working for a while, but that > somewhere along the line I'm getting a segfault, which may in turn be > triggered by a problem loading a dylib. > > I'm guessing that part of the problem is that there are a lot of moving > parts here: > > 1. RDKIT -- I'm using the latest git head as of today > 2. Python > a. version 2 or version 3? > - I'm trying to get things working with python2, but python3 would be > nice at some point > b. system installed, hombrew, or self-installed? > - I'm trying to use the homebrew-installed python2. Perhaps this is a > mistake. > 3. Python libraries (installed via pip) > 4. Homebrew libraries > 5. System libraries > 6. Environment variables > 7. etc... > > What a mess... > > speaking of environment variables, one thing that might be problematic for > me is that I am trying to use the following environment variables: > > export PYTHONPATH=/Users/sly/projects/rdkit/ > export RDBASE=/Users/sly/projects/rdkit/ > export > DYLD_LIBRARY_PATH=/System/Library/Frameworks/ImageIO.framework/Versions/A/Resources:/Users/sly/projects/rdkit/lib:/usr/local/lib: > > If I don't add the system ImageIO path, cmake fails with: > > dyld: Symbol not found: __cg_png_create_info_struct > Referenced from: > /System/Library/Frameworks/ImageIO.framework/Versions/A/ImageIO > Expected in: /usr/local/lib/libPng.dylib > in /System/Library/Frameworks/ImageIO.framework/Versions/A/ImageIO > > Adding that to DYLD_LIBRARY_PATH at least gets things to build and has > some tests running. > > Happy to provide more info if that will help figure out what's going on > here. > > thanks, > > Cyrus > > p.s. sorry about the duplicate subscription and sending this from > not-my-normal-email account, but I'm having travel-based connectivity > issues. > > |