Add solar.py back to TESTS #34
Conversation
This reverts commit 61c1500.
|
retest this please |
|
That's interesting that some machines are not robust enough to handle it |
|
@galbramc, analysis of this issue has moved to this thread |
|
What has moved here? |
|
Oh, discussion from convexengineering/gpkit#1306 Is it macys_VM, reynolds, and windows7x64 that have MOSEK 7? The new QPROP model in |
|
Yeah those are the machines with mosek 7. Can you download 7 and give it a try on your machine? Do they still support 7? |
|
it's still available for download but does not seem to be actively supported |
|
Your call then if you support 7... |
|
Don't forget that some people deal with locked down systems. Working with government agencies in particular have lots or restrictions on what they can put on their computer, and upgrading for 7 to 8 may be non-trivial. |
|
test this please |
|
@galbramc after thinking about it, and encountering another MOSEK 7/8 difference on #1315, I think we want to have MOSEK 8 on all machines. When that's done solar.py should be able to pass tests on those three machines again and we should be able to merge this. |
|
test this please |
|
@mjburton11 there appears to be a bug on this branch? it's not passing on any of the machines. |
|
Can you do anything in your build script to check that people are only using mosek 8 or newer in that case? I don't know when 9 will come out, but I'm sure it will some day. |
|
I'm going to wait until the bug on this branch is fixed to double confirm that you want the switch to mosek 8 on all machines. |
Reverts #33
@mjburton11 It only fails on the
Npods=5solve on those machines, so in hindsight I could have commented that out instead of removing the whole file from TESTS.