Closed cHYzZQo closed 6 years ago
Running locally I get the same test failures on master as travis shows on this branch. Are the failures a known issue? Or caused by the setup changes?
Thanks for working on this! It looks good, the test failures appear to be caused by the blas / lapack libraries not getting linked in correctly, which seems to suggest that the finalize_options
method is not doing what it's meant to. I tried to fix it but I couldn't figure out how to get the extra options created there to be used. Locally, you can see this by running nosetests -s
, which doesn't capture stdout, and you will see a lot of error messages like:
F----------------------------------------------------------------------------
SCS v2.0.3 - Splitting Conic Solver
(c) Brendan O'Donoghue, Stanford University, 2012-2017
----------------------------------------------------------------------------
Lin-sys: sparse-direct, nnz in A = 9075
eps = 1.00e-05, alpha = 1.50, max_iters = 100000, normalize = 1, scale = 1.00
acceleration_lookback = -1, rho_x = 1.00e-03
Variables n = 55, constraints m = 165
Cones: primal zero / dual free vars: 10
linear vars: 25
soc vars: 16, soc blks: 4
sd vars: 99, sd blks: 7
primal + dual power vars: 15
FATAL: Cannot solve SDPs with > 2x2 matrices without linked blas+lapack libraries
Install blas+lapack and re-compile SCS with blas+lapack libray locations
ERROR: init_cone failure
Setup time: 2.97e-04s
Failure:could not initialize work
Hmm, that means that when i build master locally I'm also not linking in correctly. Does master work for you?
Master works for me, yes.
Got the tests to pass by moving the list settings down to the build_extension level.
Awesome, thanks!
Thanks for merging. Any idea when this will be available in pypi?
Up now.
Allows this package to bring it in as a dep instead of forcing it to be installed already.