- cctbxbb - phenix-online.org

The curious incident of the test at install time
by markus.gerstel＠diamond.ac.uk 13 May '16

13 May '16

Hi everyone, we have observed the test cctbx_project/cctbx/regression/tst_grm_pickling.py to be failing, however only on an installation installed from the installer. It appears to run successfully on our bootstrapped developer build. I presume that the test requires some external data that we do not distribute? The relevant change occurred between r24437 and r24441, and is almost certainly r24439, which enabled the failing bit of the test. Stacktrace below. -Markus $ libtbx.python /dls/science/groups/scisoft/DIALS/CD/nightly/dials-dev20160513/modules/cctbx_project/cctbx/regression/tst_grm_pickling.py Time pickling/unpickling: 0.0187, 0.0097 Traceback (most recent call last): File "/dls/science/groups/scisoft/DIALS/CD/nightly/dials-dev20160513/modules/cctbx_project/cctbx/regression/tst_grm_pickling.py", line 198, in <module> exercise_all(sys.argv[1:]) File "/dls/science/groups/scisoft/DIALS/CD/nightly/dials-dev20160513/modules/cctbx_project/cctbx/regression/tst_grm_pickling.py", line 195, in exercise_all test_nucleic_acid(mon_lib_srv, ener_lib) File "/dls/science/groups/scisoft/DIALS/CD/nightly/dials-dev20160513/modules/cctbx_project/cctbx/regression/tst_grm_pickling.py", line 177, in test_nucleic_acid log=null_out()) File "/dls/science/groups/scisoft/DIALS/CD/nightly/dials-dev20160513/modules/cctbx_project/mmtbx/monomer_library/pdb_interpretation.py", line 5455, in run log=log) File "/dls/science/groups/scisoft/DIALS/CD/nightly/dials-dev20160513/modules/cctbx_project/mmtbx/monomer_library/pdb_interpretation.py", line 5005, in __init__ restraints_loading_flags=restraints_loading_flags, File "/dls/science/groups/scisoft/DIALS/CD/nightly/dials-dev20160513/modules/cctbx_project/mmtbx/monomer_library/pdb_interpretation.py", line 2956, in __init__ log=log, File "/dls/science/groups/scisoft/DIALS/CD/nightly/dials-dev20160513/modules/cctbx_project/mmtbx/monomer_library/pdb_interpretation.py", line 2343, in __init__ m_j=mm) File "/dls/science/groups/scisoft/DIALS/CD/nightly/dials-dev20160513/modules/cctbx_project/mmtbx/monomer_library/pdb_interpretation.py", line 1568, in get_lib_link return mon_lib_srv.link_link_id_dict["rna3p"] KeyError: 'rna3p' -- This e-mail and any attachments may contain confidential, copyright and or privileged material, and are for the use of the intended addressee only. If you are not the intended addressee or an authorised recipient of the addressee please notify us of receipt by returning the e-mail and do not use, copy, retain, distribute or disclose the information in or attached to the e-mail. Any opinions expressed within this e-mail are those of the individual and not necessarily of Diamond Light Source Ltd. Diamond Light Source Ltd. cannot guarantee that this e-mail or any attachments are free from viruses and we cannot accept liability for any damage which you may sustain as a result of software viruses which may be transmitted in or with the message. Diamond Light Source Limited (company no. 4375679). Registered in England and Wales with its registered office at Diamond House, Harwell Science and Innovation Campus, Didcot, Oxfordshire, OX11 0DE, United Kingdom

2 1

easy_mp stack traces
by Graeme.Winter＠diamond.ac.uk 13 May '16

13 May '16

Dear Gabor, some time back Markus added the capability to retain stack traces through easy_mp execution. These were reverted which now means that by default the stack traces are lost :( I don't have a clear understanding what the exact design of the newly built scheduling/stacktrace system is, perhaps you could explain this since we’re in the dark here. We are however pretty certain that the corner cases that lead to you modifying the code do not apply to us with DIALS or xia2. Could you please advise how we can now enable the stacktrace to be preserved globally, in all cases, and for all dials.*/xia2.* programs? Just today this has come up twice, and is a support issue for us, see e.g. https://github.com/dials/dials/issues/139 where we are simply flying blind. Thanks in advance and best wishes, Graeme & xia2 / DIALS team -- This e-mail and any attachments may contain confidential, copyright and or privileged material, and are for the use of the intended addressee only. If you are not the intended addressee or an authorised recipient of the addressee please notify us of receipt by returning the e-mail and do not use, copy, retain, distribute or disclose the information in or attached to the e-mail. Any opinions expressed within this e-mail are those of the individual and not necessarily of Diamond Light Source Ltd. Diamond Light Source Ltd. cannot guarantee that this e-mail or any attachments are free from viruses and we cannot accept liability for any damage which you may sustain as a result of software viruses which may be transmitted in or with the message. Diamond Light Source Limited (company no. 4375679). Registered in England and Wales with its registered office at Diamond House, Harwell Science and Innovation Campus, Didcot, Oxfordshire, OX11 0DE, United Kingdom

2 1

Installation with distutils?
by Luc Bourhis 09 May '16

09 May '16

Hi fellow cctbx developers, unless I am missing something, nobody has written a “setup.py” to install the cctbx with distutils in Python standard site-packages? If so, I am planning to contribute one as this would make it easier for us at Bruker. I am just asking in case I would duplicate existing work. Best wishes, Luc

3 4

problem when using Python to convert from numpy array to flex.int array
by Jun Aishima 04 May '16

04 May '16

I have used the bootstrap method to build cctbx with my own Python (in a virtual environment). I setup my virtual environment, then source $CCTBX/build/setpaths.sh (bash). I am currently trying to turn a detector image (uint16) read in with fabio, which results in a numpy ndarray, and then turn that into a scitbx flex array. Since I've had problems doing this in my system, I've written a little script that reproduces the problem: ********** import fabio file=fabio.open('testcrystal_1_001.img') # adsc detector image file.data[0] from scitbx.array_family import flex flex.int(file.data[3071]) # this is a 3072x3072 image. For some reason, row 0 causes no problems. ********** At the end of this email is the stack trace I get. Has anybody had similar issues? I first saw this when I was trying to convert the entire image into a flex array, but now I've seen it with only one row [3071], while still working for [0:1500], so it doesn't seem like a simple memory issue. Any help would be appreciated. Jun Aishima ********** show_stack(1): <stdin>(1) <module> libc backtrace (23 frames, most recent call last): /home/aishimaj/spotter/bin/python() [0x414199] /lib64/libc.so.6(__libc_start_main+0xfd) [0x366d21ed5d] /home/aishimaj/spotter/bin/python(Py_Main+0xb22) [0x4150b2] /home/aishimaj/spotter/bin/python(PyRun_AnyFileExFlags+0x6b) [0x4ce82b] /home/aishimaj/spotter/bin/python(PyRun_InteractiveLoopFlags+0x4e) [0x4ce01e] /home/aishimaj/spotter/bin/python(PyRun_InteractiveOneFlags+0x1ff) [0x4cdd9f] /home/aishimaj/spotter/bin/python(PyEval_EvalCode+0x32) [0x4ad362] /home/aishimaj/spotter/bin/python(PyEval_EvalCodeEx+0x877) [0x4ad257] /home/aishimaj/spotter/bin/python(PyEval_EvalFrameEx+0x6a0b) [0x4aa8bb] /home/aishimaj/spotter/bin/python(PyObject_Call+0x47) [0x419ba7] /home/aishimaj/spotter/bin/python() [0x46df9f] /home/aishimaj/spotter/bin/python() [0x47623f] /home/aishimaj/spotter/bin/python(PyObject_Call+0x47) [0x419ba7] /home/aishimaj/spotter/bin/python() [0x42296f] /home/aishimaj/spotter/bin/python(PyObject_Call+0x47) [0x419ba7] /beamline/phenix/bootstrap/build/lib/libboost_python.so(+0x2d088) [0x7f5561101088] /beamline/phenix/bootstrap/build/lib/libboost_python.so(boost::python::handle_exception_impl(boost::function0<void>)+0x43) [0x7f5561112013] /beamline/phenix/bootstrap/build/lib/libboost_python.so(+0x2fe08) [0x7f5561103e08] /beamline/phenix/bootstrap/build/lib/libboost_python.so(boost::python::objects::function::call(_object*, _object*) const+0xef) [0x7f55611031ff] /beamline/phenix/bootstrap/build/lib/scitbx_array_family_flex_ext.so(boost::python::objects::signature_py_function_impl<boost::python::detail::caller<scitbx::af::versa<int, scitbx::af::flex_grid<scitbx::af::small<long, 10ul> > >* (*)(boost::python::numeric::array const&), boost::python::detail::constructor_policy<boost::python::default_call_policies>, boost::mpl::vector2<scitbx::af::versa<int, scitbx::af::flex_grid<scitbx::af::small<long, 10ul> > >*, boost::python::numeric::array const&> >, boost::mpl::v_item<void, boost::mpl::v_item<boost::python::api::object, boost::mpl::v_mask<boost::mpl::vector2<scitbx::af::versa<int, scitbx::af::flex_grid<scitbx::af::small<long, 10ul> > >*, boost::python::numeric::array const&>, 1>, 1>, 1> >::operator()(_object*, _object*)+0x52) [0x7f555ed68452] /beamline/phenix/bootstrap/build/lib/scitbx_array_family_flex_ext.so(scitbx::af::boost_python::flex_int_from_numpy_array(boost::python::numeric::array const&)+0x29) [0x7f555f039f79] /beamline/phenix/bootstrap/build/lib/scitbx_array_family_flex_ext.so(scitbx::af::versa<int, scitbx::af::flex_grid<scitbx::af::small<long, 10ul> > > scitbx::af::boost_python::versa_flex_from_numpy_array<int>(boost::python::numeric::array const&)+0x1c87) [0x7f555f0415c7] /lib64/libc.so.6() [0x366d2326a0] Segmentation fault (Python and libc call stacks above) This crash may be due to a problem in any imported Python module, including modules which are not part of the cctbx project. To disable the traps leading to this message, define these environment variables (e.g. assign the value 1): BOOST_ADAPTBX_FPE_DEFAULT BOOST_ADAPTBX_SIGNALS_DEFAULT This will NOT solve the problem, just mask it, but may allow you to proceed in case it is not critical. Jun Aishima | Post Doc - MX | Australian Synchrotron p: (03) 8540 4114 | f: (03) 8540 4200 jun.aishima(a)synchrotron.org.au<mailto:[email protected]> | www.synchrotron.org.au 800 Blackburn Road, Clayton, Victoria 3168

3 6

Algorithm developer postdoc opening at Berkeley National Lab
by Nicholas Sauter 29 Apr '16

29 Apr '16

Berkeley Lab's Molecular Biophysics and Integrated Bioimaging Division <http://biosciences.lbl.gov/divisions/mbib/> is looking for a Computational Biologist Postdoctoral Fellow to work in the Sauter group. We are seeking an algorithm developer for the increasingly complex analysis of large diffraction datasets in structural biology. Current projects utilize XFEL crystallography and spectroscopy to investigate the photosynthetic mechanism of water splitting and to probe other metalloenzyme reactions. We also wish to test whether diffuse scattering can reveal correlated atomic motions in crystals. Many problems remain to be solved, including the details of how to optimally merge datasets from thousands of crystals. Our software development projects (including packages such as DIALS <http://dials.lbl.gov> and cctbx.xfel <http://cci.lbl.gov/xfel>) have been highlighted in several high-impact publications listed HERE <http://biosciences.lbl.gov/profiles/nicholas-sauter>. Candidates should have expertise in one or more computational techniques including, but not limited to, ray tracing to test underlying physical models of the diffraction, Bayesian approaches for refining model parameters, macromolecular modeling and refinement, neural networks for interpreting image features, as well as signal processing and denoising methods. Extensive data analysis experience in crystallography or from more general bioimaging backgrounds are welcome. We particularly encourage strong mathematical intuition and a track record of bringing new ideas and tools to fruition, as evidenced through written publication and clear presentation. Candidates should send an expression of interest, CV, and list of three references to *Nick Sauter, nksauter(a)lbl.gov <nksauter(a)lbl.gov>.* Further details are posted at https://lbl.taleo.net/careersection/2/jobdetail.ftl?lang=en&job=82354. Nicholas K. Sauter, Ph. D. Computer Staff Scientist, Molecular Biophysics and Integrated Bioimaging Division Lawrence Berkeley National Laboratory 1 Cyclotron Rd., Bldg. 33R0345 Berkeley, CA 94720

1 0

Re: [cctbxbb] new fail on windows
by R. D. Oeffner 25 Apr '16

25 Apr '16

Markus, I'd say that this is a wrong take on the current state of affairs regarding supporting Windows. I believe line 22 in easy_mp.py you refer to is supposed to be one of the last warnings one would get if one were to use some of the functions there with Windows in an unsupported manner, say if objects can't be pickled. I don't think it means this file was written with non-windows support intentionally. I appreciate people code on different platforms but the way of working shouldn't be aloof denial of the existence of Windows. In the short term until you get a Windows PC perhaps Billy Poon might grant you access to one of the virtual Windows systems he has set up for testing. Otherwise I'm happy to run tests for you every now and then on my PC. Regards, Rob -----Original Message----- From: cctbxbb-request(a)phenix-online.org Sent: Saturday, April 23, 2016 20:00 To: cctbxbb(a)phenix-online.org Subject: cctbxbb Digest, Vol 85, Issue 23 Send cctbxbb mailing list submissions to cctbxbb(a)phenix-online.org To subscribe or unsubscribe via the World Wide Web, visit http://phenix-online.org/mailman/listinfo/cctbxbb or, via email, send a message with subject or body 'help' to cctbxbb-request(a)phenix-online.org You can reach the person managing the list at cctbxbb-owner(a)phenix-online.org When replying, please edit your Subject line so it is more specific than "Re: Contents of cctbxbb digest..." Today's Topics: 1. Re: new fail on windows (markus.gerstel(a)diamond.ac.uk) ---------------------------------------------------------------------- Message: 1 Date: Fri, 22 Apr 2016 21:26:38 +0000 From: <markus.gerstel(a)diamond.ac.uk> To: <cctbxbb(a)phenix-online.org> Subject: Re: [cctbxbb] new fail on windows Message-ID: <415F7F573406BB498256C68158D456746125CE10(a)EXCHMBX01.fed.cclrc.ac.uk> Content-Type: text/plain; charset=iso-8859-1 Hi Nigel, Should be resolved now. It is true that we do not currently test on windows. In this particular case it turns out neither does anyone else, cf. https://sourceforge.net/p/cctbx/code/HEAD/tree/trunk/libtbx/easy_mp.py line 22 which causes the other easy_mp tests to be skipped on Windows. This will make it kind of difficult to support windows software building on top of it, so at some point this needs to be addressed properly. Best wishes & good weekend -Markus ________________________________ From: cctbxbb-bounces(a)phenix-online.org [cctbxbb-bounces(a)phenix-online.org] on behalf of Nigel Moriarty [nwmoriarty(a)lbl.gov] Sent: Friday, April 22, 2016 17:39 To: cctbx mailing list Subject: Re: [cctbxbb] new fail on windows On Fri, Apr 22, 2016 at 1:35 AM, <markus.gerstel(a)diamond.ac.uk<mailto:[email protected]>> wrote: Hopefully this is fixed with r24353. Afraid not. Is it true that you are not testing this windows feature on a windows machine? Regarding junit, I'll talk with Aaron when he returns. Cheers Nigel --- Nigel W. Moriarty Building 33R0349, Molecular Biophysics and Integrated Bioimaging Lawrence Berkeley National Laboratory Berkeley, CA 94720-8235 Phone : 510-486-5709 Email : NWMoriarty(a)LBL.gov Fax : 510-486-5909 Web : CCI.LBL.gov<http://CCI.LBL.gov> -- This e-mail and any attachments may contain confidential, copyright and or privileged material, and are for the use of the intended addressee only. If you are not the intended addressee or an authorised recipient of the addressee please notify us of receipt by returning the e-mail and do not use, copy, retain, distribute or disclose the information in or attached to the e-mail. Any opinions expressed within this e-mail are those of the individual and not necessarily of Diamond Light Source Ltd. Diamond Light Source Ltd. cannot guarantee that this e-mail or any attachments are free from viruses and we cannot accept liability for any damage which you may sustain as a result of software viruses which may be transmitted in or with the message. Diamond Light Source Limited (company no. 4375679). Registered in England and Wales with its registered office at Diamond House, Harwell Science and Innovation Campus, Didcot, Oxfordshire, OX11 0DE, United Kingdom ------------------------------ _______________________________________________ cctbxbb mailing list cctbxbb(a)phenix-online.org http://phenix-online.org/mailman/listinfo/cctbxbb End of cctbxbb Digest, Vol 85, Issue 23 ***************************************

2 1

new fail on windows
by Nigel Moriarty 24 Apr '16

24 Apr '16

******************************************************************************** ERROR: 1 TEST FAILURES. PLEASE FIX BEFORE COMMITTING CODE. ******************************************************************************** Standard error: Traceback (most recent call last): File "c:\users\builder\slave\cctbx-nightly-intel-windows\modules\cctbx_project\libtbx\tst_easy_mp.py", line 210, in <module> run(args=sys.argv[1:]) File "c:\users\builder\slave\cctbx-nightly-intel-windows\modules\cctbx_project\libtbx\tst_easy_mp.py", line 205, in run check_if_stacktrace_is_propagated_properly(method='multiprocessing') File "c:\users\builder\slave\cctbx-nightly-intel-windows\modules\cctbx_project\libtbx\tst_easy_mp.py", line 187, in check_if_stacktrace_is_propagated_properly assert isinstance(e, exceptions.ZeroDivisionError) or "ZeroDivisionError" in str(e), "Exception type mismatch: %s" % str(e) AssertionError: Exception type mismatch: Can't pickle <function may_divide_by_zero at 0x01C06470>: it's not found as __main__.may_divide_by_zero Traceback (most recent call last): File "<string>", line 1, in <module> File "C:\Users\builder\slave\cctbx-nightly-intel-windows\base\bin\python\lib\multiprocessing\forking.py", line 381, in main self = load(from_parent) File "C:\Users\builder\slave\cctbx-nightly-intel-windows\base\bin\python\lib\pickle.py", line 1378, in load return Unpickler(file).load() File "C:\Users\builder\slave\cctbx-nightly-intel-windows\base\bin\python\lib\pickle.py", line 858, in load dispatch[key](self) File "C:\Users\builder\slave\cctbx-nightly-intel-windows\base\bin\python\lib\pickle.py", line 880, in load_eof raise EOFError EOFError Cheers Nigel --- Nigel W. Moriarty Building 33R0349, Molecular Biophysics and Integrated Bioimaging Lawrence Berkeley National Laboratory Berkeley, CA 94720-8235 Phone : 510-486-5709 Email : NWMoriarty(a)LBL.gov Fax : 510-486-5909 Web : CCI.LBL.gov

3 4

Re: [cctbxbb] cctbxbb Digest, Vol 85, Issue 21
by R.D. Oeffner 22 Apr '16

22 Apr '16

Hi Markus, I was gonna ask why don't you guys set up a windows machine and test for the cctbx crash on that? The VC++9.0 compiler is free and if you use the python that comes in a phenix for windows installation the necessary modules are all present. Rob On 22/04/2016 09:35, cctbxbb-request(a)phenix-online.org wrote: > Send cctbxbb mailing list submissions to > cctbxbb(a)phenix-online.org > > To subscribe or unsubscribe via the World Wide Web, visit > http://phenix-online.org/mailman/listinfo/cctbxbb > or, via email, send a message with subject or body 'help' to > cctbxbb-request(a)phenix-online.org > > You can reach the person managing the list at > cctbxbb-owner(a)phenix-online.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of cctbxbb digest..." > > > Today's Topics: > > 1. Re: new fail on windows (markus.gerstel(a)diamond.ac.uk) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Fri, 22 Apr 2016 08:35:42 +0000 > From: <markus.gerstel(a)diamond.ac.uk> > To: <cctbxbb(a)phenix-online.org> > Subject: Re: [cctbxbb] new fail on windows > Message-ID: > <415F7F573406BB498256C68158D456746125C23B(a)EXCHMBX01.fed.cclrc.ac.uk> > Content-Type: text/plain; charset="utf-8" > > Hi Nigel, > > Hopefully this is fixed with r24353. > > One question: Do you think it would be possible to install > https://pypi.python.org/pypi/junit-xml on your build machines, run > libtbx.run_tests_parallel with the output_junit_xml=True option, and > collect the output.xml files somewhere where we can download them? > > At the moment we do monitor the buildbot status - but only in a very > coarse way. We only see test failures on a high level, see attached > screenshot. As you can see there are a number of things broken, some > of them for quite some time, so because on that level there is always > something guaranteed to fail the overall status we see is always set > to ?warning?. I don?t think any of us bothers anymore looking into the > details, a number of errors are about things we can?t do anything > about anyway. > > However, if you could give us a collection of output.xml?s we could, > instead of taking the overall build status, use the test results and > set up a DIALS-East internal mail on the buildbot test status. Then we > would pick up these things earlier. > > Best wishes, > -Markus > > From: cctbxbb-bounces(a)phenix-online.org > [mailto:[email protected]] On Behalf Of Nigel Moriarty > Sent: 21 April 2016 17:41 > To: cctbx mailing list > Subject: [cctbxbb] new fail on windows > > > ******************************************************************************** > > ERROR: 1 TEST FAILURES. PLEASE FIX BEFORE COMMITTING CODE. > > ******************************************************************************** > > > > Standard error: > > Traceback (most recent call last): > > File > "c:\users\builder\slave\cctbx-nightly-intel-windows\modules\cctbx_project\libtbx\tst_easy_mp.py", > line 210, in <module> > > run(args=sys.argv[1:]) > > File > "c:\users\builder\slave\cctbx-nightly-intel-windows\modules\cctbx_project\libtbx\tst_easy_mp.py", > line 205, in run > > > check_if_stacktrace_is_propagated_properly(method='multiprocessing') > > File > "c:\users\builder\slave\cctbx-nightly-intel-windows\modules\cctbx_project\libtbx\tst_easy_mp.py", > line 187, in check_if_stacktrace_is_propagated_properly > > assert isinstance(e, exceptions.ZeroDivisionError) or > "ZeroDivisionError" in str(e), "Exception type mismatch: %s" % str(e) > > AssertionError: Exception type mismatch: Can't pickle <function > may_divide_by_zero at 0x01C06470>: it's not found as > __main__.may_divide_by_zero > > Traceback (most recent call last): > > File "<string>", line 1, in <module> > > File > "C:\Users\builder\slave\cctbx-nightly-intel-windows\base\bin\python\lib\multiprocessing\forking.py", > line 381, in main > > self = load(from_parent) > > File > "C:\Users\builder\slave\cctbx-nightly-intel-windows\base\bin\python\lib\pickle.py", > line 1378, in load > > return Unpickler(file).load() > > File > "C:\Users\builder\slave\cctbx-nightly-intel-windows\base\bin\python\lib\pickle.py", > line 858, in load > > dispatch[key](self) > > File > "C:\Users\builder\slave\cctbx-nightly-intel-windows\base\bin\python\lib\pickle.py", > line 880, in load_eof > > raise EOFError > > EOFError > Cheers > > Nigel > > --- > Nigel W. Moriarty > Building 33R0349, Molecular Biophysics and Integrated Bioimaging > Lawrence Berkeley National Laboratory > Berkeley, CA 94720-8235 > Phone : 510-486-5709 Email : > NWMoriarty(a)LBL.gov<mailto:[email protected]> > Fax : 510-486-5909 Web : CCI.LBL.gov<http://CCI.LBL.gov> > > -- > This e-mail and any attachments may contain confidential, copyright > and or privileged material, and are for the use of the intended > addressee only. If you are not the intended addressee or an authorised > recipient of the addressee please notify us of receipt by returning > the e-mail and do not use, copy, retain, distribute or disclose the > information in or attached to the e-mail. > Any opinions expressed within this e-mail are those of the individual > and not necessarily of Diamond Light Source Ltd. > Diamond Light Source Ltd. cannot guarantee that this e-mail or any > attachments are free from viruses and we cannot accept liability for > any damage which you may sustain as a result of software viruses which > may be transmitted in or with the message. > Diamond Light Source Limited (company no. 4375679). Registered in > England and Wales with its registered office at Diamond House, Harwell > Science and Innovation Campus, Didcot, Oxfordshire, OX11 0DE, United > Kingdom >

2 1

Trapping NaN or not trapping NaN
by Luc Bourhis 14 Apr '16

14 Apr '16

Richard and I have been hammering at our build system and code to try to solve several crashes on OS X with Xcode 7.3. Several of them came from the fact that we trap NaN whereas Xcode 7.3 aggressive optimisations produce code that may create a NaN but still deliver the correct final result. A trivial example came from cctbx/xray/conversions.h. We have if (f_sq > 0) f = std::sqrt(f_sq); else f = 0; With the version of clang shipping with Xcode 7.3, both branches are evaluated in parallel before selecting the result based on the test f_sq > 0. This is done so as not to stall the pipeline and this can result in much faster code indeed. But that means that every time f_sq < 0, a NaN will be produced, and in the context of our tests, this means we crash. Richard and I avoided the elephant in the room, stop to trap NaN, and as a result came with uglier and uglier fixes. Moreover, we are now at the mercy of the compiler applying such an optimisation to another file at the next update of Xcode, which would then force us to fix the compilation of said file in the relevant SConscript. It’s a loosing game. Thus I reckon that at least on OS X, we should stop trapping NaN. That will leave us with the true clang regression for which I will try to submit reduced cases to the clang bugzilla. Best wishes, Luc

5 7

Boost broken on mac?
by markus.gerstel＠diamond.ac.uk 12 Apr '16

12 Apr '16

Dials west is currently observing test failures on Mac related to Boost. To exclude any errors on our side I have run things twice clean from bootstrap: One of a number of affected tests: libtbx.python cctbx/xray/boost_python/tst_xray.py show_stack(1): /Users/dlshudson/jenkins_slave/workspace/dials_package_macosx/build_dials/modules/cctbx_project/cctbx/xray/boost_python/tst_xray.py(381) exercise_conversions show_stack(2): /Users/dlshudson/jenkins_slave/workspace/dials_package_macosx/build_dials/modules/cctbx_project/cctbx/xray/boost_python/tst_xray.py(1993) run show_stack(3): /Users/dlshudson/jenkins_slave/workspace/dials_package_macosx/build_dials/modules/cctbx_project/cctbx/xray/boost_python/tst_xray.py(2009) <module> libc backtrace (29 frames, most recent call last): 30 Python 0x0000000100000d71 Python + 3441 29 Python 0x0000000100000e58 Python + 3672 28 Python 0x00000001000ef06f Py_Main + 3135 27 Python 0x00000001000d78be PyRun_SimpleFileExFlags + 702 26 Python 0x00000001000d7d84 PyRun_FileExFlags + 164 25 Python 0x00000001000aea06 PyEval_EvalCode + 54 24 Python 0x00000001000af244 PyEval_EvalCodeEx + 2100 23 Python 0x00000001000b25c8 PyEval_EvalFrameEx + 12280 22 Python 0x00000001000b6ab6 fast_function + 118 21 Python 0x00000001000af244 PyEval_EvalCodeEx + 2100 20 Python 0x00000001000b25c8 PyEval_EvalFrameEx + 12280 19 Python 0x00000001000b6ab6 fast_function + 118 18 Python 0x00000001000af244 PyEval_EvalCodeEx + 2100 17 Python 0x00000001000b4198 PyEval_EvalFrameEx + 19400 16 Python 0x0000000100011023 PyObject_Call + 99 15 Python 0x000000010006ad8b type_call + 347 14 Python 0x00000001000708af slot_tp_init + 175 13 Python 0x0000000100011023 PyObject_Call + 99 12 Python 0x000000010001f0d8 instancemethod_call + 232 11 Python 0x0000000100011023 PyObject_Call + 99 10 libboost_python.dylib 0x0000000101888663 boost::python::objects::function_call(_object*, _object*, _object*) + 83 9 libboost_python.dylib 0x000000010188ce71 boost::python::handle_exception_impl(boost::function0<void>) + 81 8 libboost_python.dylib 0x00000001018889ca boost::detail::function::void_function_ref_invoker0<boost::python::objects::(anonymous namespace)::bind_return, void>::invoke(boost::detail::function::function_buffer&) + 26 7 libboost_python.dylib 0x000000010188693a boost::python::objects::function::call(_object*, _object*) const + 874 6 cctbx_xray_ext.so 0x0000000107c8e1c5 boost::python::detail::caller_arity<3u>::impl<void (*)(_object*, scitbx::af::const_ref<double, scitbx::af::trivial_accessor> const&, scitbx::af::const_ref<double, scitbx::af::trivial_accessor> const&), boost::python::default_call_policies, boost::mpl::vector4<void, _object*, scitbx::af::const_ref<double, scitbx::af::trivial_accessor> const&, scitbx::af::const_ref<double, scitbx::af::trivial_accessor> const&> >::operator()(_object*, _object*) + 229 5 cctbx_xray_ext.so 0x0000000107c91f44 boost::python::objects::make_holder<2>::apply<boost::python::objects::value_holder<cctbx::xray::array_f_sq_as_f<cctbx::xray::f_sq_as_f_xtal_3_7, double> >, boost::mpl::joint_view<boost::python::detail::drop1<boost::python::detail::type_list<scitbx::af::const_ref<double, scitbx::af::trivial_accessor> const&, scitbx::af::const_ref<double, scitbx::af::trivial_accessor> const&, boost::python::optional<double const&, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_>, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_> >, boost::python::optional<double const&, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_, mpl_::void_> > >::execute(_object*, scitbx::af::const_ref<double, scitbx::af::trivial_accessor> const&, scitbx::af::const_ref<double, scitbx::af::trivial_accessor> const&) + 100 4 ??? 0x0000000000000002 0x0 + 2 3 libsystem_platform.dylib 0x00007fff8caabeaa _sigtramp + 26 2 boost_python_meta_ext.so 0x0000000101841370 initboost_python_meta_ext + 0 Floating-point error (Python and libc call stacks above) This crash may be due to a problem in any imported Python module, including modules which are not part of the cctbx project. To disable the traps leading to this message, define these environment variables (e.g. assign the value 1): BOOST_ADAPTBX_FPE_DEFAULT BOOST_ADAPTBX_SIGNALS_DEFAULT This will NOT solve the problem, just mask it, but may allow you to proceed in case it is not critical. Any ideas? Best wishes, -Markus -- This e-mail and any attachments may contain confidential, copyright and or privileged material, and are for the use of the intended addressee only. If you are not the intended addressee or an authorised recipient of the addressee please notify us of receipt by returning the e-mail and do not use, copy, retain, distribute or disclose the information in or attached to the e-mail. Any opinions expressed within this e-mail are those of the individual and not necessarily of Diamond Light Source Ltd. Diamond Light Source Ltd. cannot guarantee that this e-mail or any attachments are free from viruses and we cannot accept liability for any damage which you may sustain as a result of software viruses which may be transmitted in or with the message. Diamond Light Source Limited (company no. 4375679). Registered in England and Wales with its registered office at Diamond House, Harwell Science and Innovation Campus, Didcot, Oxfordshire, OX11 0DE, United Kingdom

6 35