[release-0.5] backports for 0.5.1 #20736
Conversation
This is the 0.5 edition: with this change a, b, n are always hit
exactly by a:s:a+n*s construction tests. There are two parts to
the fix: 1) check that the lifted rational endpoint is exact even
*after* we've reduced terms of the fractions; 2) apply the same
fix that I've proposed for the length in the non-lifted case.
Combined backport of these commits:
- e849169
- b7ad743
range tests: allow any range length that hits stop (#20532)
(cherry picked from commit b7ad743)
) (cherry picked from commit 2b2894c)
* lufact for sparse matrix pivot option error
base/sparse/umfpack.jl includes the following method definition for lufact
`lufact(A::SparseMatrixCSC, pivot::Type{Val{false}}) = lufact(A)`
This should likely be
`lufact(A::SparseMatrixCSC, pivot::Type{Val{true}}) = lufact(A)`
because in lufact pivoting is on by default.
The error is shown in the following example
```
A = speye(4)
A[1:2,1:2] = [-.01 -200; 200 .001]
F = lufact(A,Val{false})
F[:p]
```
which returns
```
julia> F[:q]
4-element Array{Int64,1}:
3
4
1
2
```
However it should return
```
julia> F[:q]
4-element Array{Int64,1}:
1
2
3
4
```
because pivoting was turned off.
* Added test for #18246 and #18244
* 4 space indent
oops my bad
* remove unnecessary lufact method
* update test for removed lufact method definition
(cherry picked from commit 68d3d32)
This enables calls to kwarg functions before Base is defined (eg. in inference), as the frontend generates a call to vector_any. (cherry picked from commit 95c26b5) ref #18396 Add a test for #18396. Call a kwarg function in Core.Inference where vector_any was unavailable. (cherry picked from commit fab9560)
* Update require build tools Notes: - LLVM fails to build unless you have Python 2.7+ installed: - cmake fails to build libgit2 unless the version is at least 3.1.2. For details see: https://cmake.org/Bug/view.php?id=15386 * Fix Python link * Add build troubleshooting row * Remove CMake version requirement (cherry picked from commit a0cec60)
(cherry picked from commit 9ebf5d6)
(cherry picked from commit 329e7e1) ref #18522 include fix for #14427 (cherry picked from commit 259e174) expand README (cherry picked from commit 399710e) remove use of sudo from OS X app makefile (cherry picked from commit 55fa673) add recommended Info.plist keys (cherry picked from commit aeadb11)
I'm already tired of seeing bug reports with the output of versioninfo() that don't say Windows anywhere. Many users don't know what NT or Darwin mean. (cherry picked from commit 75d251e)
(cherry picked from commit 990dcd9)
|
@nanosoldier |
|
Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @jrevels |
tkelman
changed the title
|
pkgeval identified some regressions here that I'll have to bisect on (any help would be appreciated): https://gist.github.com/anonymous/4178d02e88176194746115ff641a21b5 |
|
@JeffBezanson the backport of #18493 broke ASTInterpreter. If you aren't aware of a fix for that then I'm going to revert f32c0ed. edit: this is also what broke FixedSizeArrays, and GeometryTypes, and QuantumOptics (though that last one has an annoying problem, fixed on QuantumOptics master, where the package was running many of its tests by shelling out to whatever version of julia happened to be on the path, not necessarily the version of julia that ran |
|
cc @denizyuret I think the failure in Knet.jl (log at https://gist.github.com/8ab4a5e8f0e56370221cbe39924e41fa) is intermittent, I can get it to happen even on 0.5.0 but it requires re-running the tests multiple times to get to happen. So don't think that's a real regression that any of the backports here caused. |
|
The failure in OrdinaryDiffEq is a wrong |
tkelman
force-pushed
the
tk/backports-0.5.1
branch
from
f6003b4 to
c2bede6
Compare
point to release-0.5 instead of master or latest
tkelman
force-pushed
the
tk/backports-0.5.1
branch
from
c2bede6 to
0cc7d79
Compare
|
merge commit for PR, currently 0cc7d79: @nanosoldier I've reverted the conversion change. |
|
Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @jrevels |
|
pkgeval results: https://gist.github.com/5b71a57370610d91495f79004f58f354 both failures, InformationMeasures.jl and Tensors.jl, are flaky tolerances. |
|
Woot! |
this is way overdue,
and still in progress making my way through the list of pending backportswill run pkgeval and nanosoldier here when ready