Merge branch 'symengine-fixes' into 'master' (f52fefb5) · Commits · Andreas Klöckner / sumpy

@mattwala @isuruf Ouch. The conda build died with a segfault, after passing in !54 (merged). Any clue what might be happening?

Test node ran out of memory?

Unlikely--porter has 64 GiB.

Plus swap.

Stack trace. Looks like something to do with loading SymEngine:

#0  0x00000000000570d6 in ?? ()
#1  0x00007fffc6cea258 in PyInit_symengine_wrapper ()
   from /home/wala1/sumpy-test/sumpy/.miniconda3/envs/testing/lib/python3.5/site-packages/symengine/lib/symengine_wrapper.cpython-35m-x86_64-linux-gnu.so
#2  0x00007ffff7a2d4a7 in _PyImport_LoadDynamicModuleWithSpec (spec=spec@entry=0x7fffc5f5e550, fp=fp@entry=0x0) at ./Python/importdl.c:154
#3  0x00007ffff7a2b45b in _imp_create_dynamic_impl (module=<optimized out>, file=<optimized out>, spec=0x7fffc5f5e550) at Python/import.c:2031
#4  _imp_create_dynamic (module=<optimized out>, args=<optimized out>) at Python/clinic/import.c.h:282
#5  0x00007ffff79740f9 in PyCFunction_Call (func=func@entry=0x7ffff6984ee8, args=args@entry=0x7fffc5f5e128, kwds=kwds@entry=0x7fffc5f69b88)
    at Objects/methodobject.c:109
#6  0x00007ffff7a0fed1 in ext_do_call (nk=<optimized out>, na=0, flags=<optimized out>, pp_stack=0x7fffffff1b20, func=0x7ffff6984ee8) at Python/ceval.c:5084
#7  PyEval_EvalFrameEx (f=f@entry=0x7fffc5f63930, throwflag=throwflag@entry=0) at Python/ceval.c:3328
#8  0x00007ffff7a11416 in _PyEval_EvalCodeWithName (_co=0x7ffff69790c0, globals=<optimized out>, locals=locals@entry=0x0, args=<optimized out>, argcount=2, 
    kws=0x7fffc9024b00, kwcount=0, defs=0x0, defcount=0, kwdefs=kwdefs@entry=0x0, closure=0x0, name=name@entry=0x7ffff69727b0, qualname=0x7ffff69727b0)
    at Python/ceval.c:4071
#9  0x00007ffff7a0e515 in fast_function (nk=<optimized out>, na=<optimized out>, n=2, pp_stack=0x7fffffff1d40, func=<optimized out>) at Python/ceval.c:4866
#10 call_function (oparg=<optimized out>, pp_stack=0x7fffffff1d40) at Python/ceval.c:4783
#11 PyEval_EvalFrameEx (f=f@entry=0x7fffc9024958, throwflag=throwflag@entry=0) at Python/ceval.c:3289
#12 0x00007ffff7a0f130 in fast_function (nk=<optimized out>, na=<optimized out>, n=2, pp_stack=0x7fffffff1ec0, func=<optimized out>) at Python/ceval.c:4856
#13 call_function (oparg=<optimized out>, pp_stack=0x7fffffff1ec0) at Python/ceval.c:4783
#14 PyEval_EvalFrameEx (f=f@entry=0x7fffc9024788, throwflag=throwflag@entry=0) at Python/ceval.c:3289
#15 0x00007ffff7a0f130 in fast_function (nk=<optimized out>, na=<optimized out>, n=1, pp_stack=0x7fffffff2040, func=<optimized out>) at Python/ceval.c:4856
#16 call_function (oparg=<optimized out>, pp_stack=0x7fffffff2040) at Python/ceval.c:4783
#17 PyEval_EvalFrameEx (f=f@entry=0x7fffc7009a48, throwflag=throwflag@entry=0) at Python/ceval.c:3289
#18 0x00007ffff7a0f130 in fast_function (nk=<optimized out>, na=<optimized out>, n=1, pp_stack=0x7fffffff21c0, func=<optimized out>) at Python/ceval.c:4856
#19 call_function (oparg=<optimized out>, pp_stack=0x7fffffff21c0) at Python/ceval.c:4783
#20 PyEval_EvalFrameEx (f=f@entry=0x12779b8, throwflag=throwflag@entry=0) at Python/ceval.c:3289
#21 0x00007ffff7a0f130 in fast_function (nk=<optimized out>, na=<optimized out>, n=2, pp_stack=0x7fffffff2340, func=<optimized out>) at Python/ceval.c:4856
#22 call_function (oparg=<optimized out>, pp_stack=0x7fffffff2340) at Python/ceval.c:4783
#23 PyEval_EvalFrameEx (f=f@entry=0x7fffc7018438, throwflag=throwflag@entry=0) at Python/ceval.c:3289
#24 0x00007ffff7a11416 in _PyEval_EvalCodeWithName (_co=0x7ffff6985420, globals=<optimized out>, locals=locals@entry=0x0, args=args@entry=0x7fffc7429860, 
    argcount=2, kws=kws@entry=0x0, kwcount=kwcount@entry=0, defs=defs@entry=0x0, defcount=defcount@entry=0, kwdefs=kwdefs@entry=0x0, 
    closure=closure@entry=0x0, name=name@entry=0x0, qualname=qualname@entry=0x0) at Python/ceval.c:4071
#25 0x00007ffff7a11508 in PyEval_EvalCodeEx (_co=<optimized out>, globals=<optimized out>, locals=locals@entry=0x0, args=args@entry=0x7fffc7429860, 
    argcount=<optimized out>, kws=kws@entry=0x0, kwcount=kwcount@entry=0, defs=defs@entry=0x0, defcount=defcount@entry=0, kwdefs=0x0, closure=0x0)
    at Python/ceval.c:4092
#26 0x00007ffff7950b06 in function_call (func=0x7ffff6992f28, arg=0x7fffc7429848, kw=0x0) at Objects/funcobject.c:627
#27 0x00007ffff791f33a in PyObject_Call (func=func@entry=0x7ffff6992f28, arg=arg@entry=0x7fffc7429848, kw=kw@entry=0x0) at Objects/abstract.c:2166
#28 0x00007ffff7920249 in _PyObject_CallMethodIdObjArgs (callable=0x7ffff6992f28, name=name@entry=0x7ffff7d76860 <PyId__find_and_load.11927>)
    at Objects/abstract.c:2424
#29 0x00007ffff7a2c808 in PyImport_ImportModuleLevelObject (name=0x7fffc7420228, given_globals=<optimized out>, locals=<optimized out>, 
    given_fromlist=<optimized out>, level=1) at Python/import.c:1595
#30 0x00007ffff7a0335f in builtin___import__ (self=<optimized out>, args=<optimized out>, kwds=<optimized out>) at Python/bltinmodule.c:213
#31 0x00007ffff7974119 in PyCFunction_Call (func=0x7ffff69d3e58, args=0x7fffc7043c50, kwds=<optimized out>) at Objects/methodobject.c:98
#32 0x00007ffff791f33a in PyObject_Call (func=func@entry=0x7ffff69d3e58, arg=arg@entry=0x7fffc7043c50, kw=<optimized out>) at Objects/abstract.c:2166
#33 0x00007ffff7a06d27 in PyEval_CallObjectWithKeywords (func=func@entry=0x7ffff69d3e58, arg=arg@entry=0x7fffc7043c50, kw=kw@entry=0x0)
    at Python/ceval.c:4633

Issue is with symengine and python-symengine being pulled from different channels. This was fixed with !54 (merged)

Good catch - it does look like the build that died pulled SymEngine from conda-forge. I am not sure why but I am assuming it has to do with gitlab.

This was fixed with !54 (merged) (merged)

This is the merge commit for !54 (merged), so if anything was fixed with that, this commit should have it...

Is the conda environment cached? If not, I don't see why conda would be doing this.

The conda environment is not cached. However, the test script is not doing anything about .condarc. Could that be causing the problem?

Something wrong with the conda solver. I was able to reproduce this locally. When some of the other packages are removed, conda will use both symengine and python-symengine from symengine. Also channel prefixing channel::package_name doesn't seem to have any effect. I just pushed conda-forge::symengine=0.3.0 and now both packages are getting pulled from symengine. Weird. Can somebody restart the build?

Thanks for looking into it. I don't know if there is a way to restart this build. The best thing to do would be to change the environment file to get rid of the symengine channel now that conda-forge has 0.3.0

!56 (merged)

Thanks to the both of you for looking into this.