[release-0.4, RFC] backports for 0.4.1 #13834

tkelman · 2015-10-31T02:52:16Z

🚂

When working with Sublime as my default editor on OS X I had set my editor to: EDITOR=/Applications/Sublime\ Text.app/Contents/SharedSupport/bin/subl Since the `ispath` function did not recognize this as a path (because contained a backslash) this resulted in my editor not being recognized as "subl" and line number support being unavailable. The section of `edit` that deals with determining the user's editor has been abstracted into a new function `editor`. Test cases have been added for `editor` along with some basic tests for `shell_split`. Note that `startswith` is used to identify alternative editor names: - Sublime Text can refer to "subl" or "sublime". - Emacs can refer to "emacs" or "emacsclient". (cherry picked from commit 53fcec2) ref #13032

Moves all the cmdline option error handling code from client.jl to repl.c. (cherry picked from commit dff1c90) ref #13245

(cherry picked from commit 51b6d24) ref #13315

(cherry picked from commit d5edd11)

(cherry picked from commit b31925c) ref #13376

(cherry picked from commit e5b32a9) ref #13408

This removes the restriction on setindex_shape_check that the index lengths must all be Int, allowing data structures to return non-Int lengths (e.g., `UInt(1):UInt(2)` or `big(1):big(2)`). Fixes #13250. (cherry picked from commit e5b3cc6) ref #13252

ref #922 and 35c5234 (cherry picked from commit 99ee2ff) ref #13431

…ies. Take two. (cherry picked from commit 127ee86) ref #13437

(cherry picked from commit fa6c977) ref #13460

(cherry picked from commit 18c9029) ref #13490

(cherry picked from commit e917b71) ref #13498

from upgrade to Git 2.x, causing failure of spawn test on appveyor also remove busybox since Git 2.x makes it unnecessary (cherry picked from commit c18d271) ref #13509

(cherry picked from commit 09c9a19)

fixes #12865 (cherry picked from commit b7697e7)

Docs attached to numbers weren't being found. Finding docs for specific methods using the syntax help?> f(::Int) wasn't being recognised due to changes in how `gen_call_with_extracted_types` generated an error when it failed. (Previously the error was thrown by the function, but is now returned as an `:(error("..."))` expression.) (cherry picked from commit 446ddaa) ref #13535

hvcat doc example was not working I changed it to one that does. (cherry picked from commit 3d132a9) ref #13566

Removed test for a non-working example and added clearer doctest example instead. (cherry picked from commit 0e498d3)

(cherry picked from commit d06e348)

as long as the value is zero (cherry picked from commit 566acb1) ref #13587

Docstrings defined before `==` is available for `Symbol` comparisons causes a build error. Fix by comparing with `===`. Ref: #13534 (comment) (cherry picked from commit d8b6a24) ref #13602

(cherry picked from commit e44938c) ref #13630

(cherry picked from commit c890375) ref #13218

Also added tests/docs for those functions and exported them (cherry picked from commit 2bf7818) ref #13637

(cherry picked from commit 658a0a8) ref #13561

(cherry picked from commit 9c1f69c) ref #13455

…lication, e.g. quaternions. (cherry picked from commit fe7bb29) ref #13640

This allows the build to proceed for AArch64 (mentioned in #10791) and others. (cherry picked from commit f244fb6) ref #13558

(cherry picked from commit 316da80) ref #13641

timholy · 2015-11-01T12:12:17Z

The Images failure looks due to a specific instance of random numbers. I'll have to investigate and likely change that test, but for your purposes you can ignore that particular error.

timholy · 2015-11-01T12:28:41Z

For your benefit, justifications for #13803 include: it's necessary for jump-dev/JuMP.jl#604 and would allow operations like

using SIUnits.ShortUnits
A = [1m 2m; -1m 5m]
v = [3m,-1m]
A*v

to become type-stable (currently it returns a non-leaftype as a mechanism to circumvent what would otherwise be an error). That said, I also recognize the risk. As I stated elsewhere, since I'm obviously biased I would prefer to defer to others in making the decision about whether to include it.

The patch list looks delicious, very much looking forward to 0.4.1 even if nothing else gets added.

(cherry picked from commit 49a01db)

fix #13810 fix #13806 fix JuliaParallel/DistributedArrays.jl#6 (cherry picked from commit 843ab66)

This mirrors html rendering. Test markdown's terminal output. (cherry picked from commit 4e3aa06) ref #13835

…792. Make fwd/bwdTriSolve! work for triagular views Add check for triangular matrices in sparse factorize (cherry picked from commit 81753cc) ref #13793

hiccup7 · 2015-11-03T21:48:41Z

Now that OpenBLAS v0.2.15 has been merged to the master branch (#13808), I vote for including it in the backports for 0.4.1. I am assuming that the OpenBLAS API stayed the same. Many bugs and performance issues have been fixed in the 7 months since the last OpenBLAS release, and they include showstoppers for me in using Julia.

tkelman · 2015-11-04T02:53:28Z

Scroll up, it's already here a3f7c1f

Though when I ran the microbenchmarks it did look like randmatstat got about 15% slower, which might be attributable to the new version of openblas?

tkelman · 2015-11-04T13:19:18Z

I was wrong to blame openblas, it was actually 1fc2d8e that was responsible for the 15% slowdown in rand_mat_stat.

DataArrays has still not been tagged to work with #13803. I could be convinced to maybe pull #13491 and #13553 in here if perf and PackageEvaluator look okay on them, but unless someone argues for them I'm inclined to leave them until the next point release.

hiccup7 · 2015-11-04T18:21:55Z

@tkelman - thanks for the clarification.

jakebolewski · 2015-11-04T19:14:04Z

Is there a rough ETA on the next tagged bugfix? Some students are asking about it in @alanedelman's class.

tkelman · 2015-11-04T19:28:55Z

This weekend probably. Asked for input and opinions above and have yet to get any other than from Tim.

On Wed, Nov 4, 2015 at 11:14 AM -0800, "Jake Bolewski" notifications@github.com wrote:
Is there a rough ETA on the next tagged bugfix, some students are asking about it in @alanedelman's class.

Reply to this email directly or view it on GitHub:
#13834 (comment)

(cherry picked from commit e93465f) ref #13872

This would previously have been an infinite loop if `length(A) == typemax(Int)` so the loop vectorizer couldn't compute a trip count. (cherry picked from commit fa89a6e) ref #13866

(cherry picked from commit 23c3139) ref #13874

This is consistent with eltype() and therefore collect(). Fixes #9261. (cherry picked from commit f379e08) ref #13903

(cherry picked from commit db1ab4a) ref #13909

(cherry picked from commit 3cf7a34)

(cherry picked from commit 50c7318)

(cherry picked from commit f0dbba6)

(cherry picked from commit bbec291)

When `macroexpand` fails due to an undefined macro in the expression return the resulting error rather than the more generic and misleading "invalid doc expression..." error message. Fixes #13905. (cherry picked from commit c4c38d1) ref #13908

tkelman · 2015-11-07T21:40:44Z

https://gist.github.com/c4ae8dfc79bdc2111f7e

Most of the failures are either intermittent things I've seen before, or odd issues that only happened on the 0.4.0 baseline run where some unrelated package gives something like LoadError: module Graphics not found in current path; removed orphaned cache file /home/vagrant/.julia/lib/v0.4/Graphics.ji.

Will merge and tag shortly unless anyone wants to advocate that anything remaining on the checklist in the first comment is must-do-now.

[release-0.4, RFC] backports for 0.4.1

omus and others added 30 commits October 30, 2015 19:25

better cmdline option err handling (fixes #12679)

192e140

Moves all the cmdline option error handling code from client.jl to repl.c. (cherry picked from commit dff1c90) ref #13245

Bugfix in vcat.

b792604

(cherry picked from commit 51b6d24) ref #13315

Test case for issue 13315.

8115bcd

(cherry picked from commit d5edd11)

add trunc methods to BigInt, fixes #13367

b5ea3e1

(cherry picked from commit b31925c) ref #13376

fix ambiguity warnings trunc(BigInt,::Float16)

600f647

(cherry picked from commit e5b32a9) ref #13408

Fix inlinedAt debug locations

efdcbf5

ref #922 and 35c5234 (cherry picked from commit 99ee2ff) ref #13431

Cleanup triangular multiplication and solves and remove redundant cop…

33e1179

…ies. Take two. (cherry picked from commit 127ee86) ref #13437

Use eachindex in generic_scale!

78d5701

(cherry picked from commit fa6c977) ref #13460

Fix bugs & add tests for triangular matrices

3b83b4b

(cherry picked from commit 18c9029) ref #13490

use color in precompilation output if parent process is using color

6e02d3c

(cherry picked from commit e917b71) ref #13498

Fix Windows Git path issues

a80e901

from upgrade to Git 2.x, causing failure of spawn test on appveyor also remove busybox since Git 2.x makes it unnecessary (cherry picked from commit c18d271) ref #13509

Fix gitconfig modification with bundled git

61bd763

(cherry picked from commit 09c9a19)

Add Libdl.dlext to documentation

f26ba6c

fixes #12865 (cherry picked from commit b7697e7)

hvcat doc example fixed

6b074d0

hvcat doc example was not working I changed it to one that does. (cherry picked from commit 3d132a9) ref #13566

Added doctest for hvcat

925ca74

Removed test for a non-working example and added clearer doctest example instead. (cherry picked from commit 0e498d3)

fixing whitespaces and generating arrary.rst doc

6446a16

(cherry picked from commit d06e348)

Allow setindex! in zero part of Triangular matrices...

fbe97b1

as long as the value is zero (cherry picked from commit 566acb1) ref #13587

Fix doc bootstrap method error.

68bae30

Docstrings defined before `==` is available for `Symbol` comparisons causes a build error. Fix by comparing with `===`. Ref: #13534 (comment) (cherry picked from commit d8b6a24) ref #13602

fix leak in RemoteRef

930d63d

(cherry picked from commit e44938c) ref #13630

adds some details to at-async and tweaks at-schedule doc

df0647e

(cherry picked from commit c890375) ref #13218

Fix bug in findmin! and findmax!

083a4ca

Also added tests/docs for those functions and exported them (cherry picked from commit 2bf7818) ref #13637

Un-exporting findmin! and findmax! for release-0.4

144fc26

Bump MAX_TYPE_DEPTH again

656f3b2

(cherry picked from commit 658a0a8) ref #13561

Prefer the parent module when serializing functions (fixes #13452)

76379cb

(cherry picked from commit 9c1f69c) ref #13455

Fix generic axpy! for vector element types without commutative multip…

085a985

…lication, e.g. quaternions. (cherry picked from commit fe7bb29) ref #13640

Define all other architectures as "not x86"

57ab2f6

This allows the build to proceed for AArch64 (mentioned in #10791) and others. (cherry picked from commit f244fb6) ref #13558

Export jl_strtod_c/jl_strtof_c from src/

94f2a50

(cherry picked from commit 316da80) ref #13641

StefanKarpinski and others added 3 commits November 1, 2015 12:56

make e734f9e permanent

db51f98

(cherry picked from commit 49a01db)

fix serialization typo

ded1380

fix #13810 fix #13806 fix JuliaParallel/DistributedArrays.jl#6 (cherry picked from commit 843ab66)

regen rst docs

d52cf62

timholy mentioned this pull request Nov 3, 2015

Testing fixes JuliaImages/Images.jl#394

Merged

hayd and others added 2 commits November 3, 2015 06:32

Hide multiple spaces when rendering markdown.

4f9cbaa

This mirrors html rendering. Test markdown's terminal output. (cherry picked from commit 4e3aa06) ref #13835

Use sparse triangular solvers for sparse triangular solves. Fixes #13…

6fa0f6e

…792. Make fwd/bwdTriSolve! work for triagular views Add check for triangular matrices in sparse factorize (cherry picked from commit 81753cc) ref #13793

amitmurthy and others added 9 commits November 7, 2015 02:31

fix parametric type definitions for RemoteRef and AbstractChannel

04f3d15

(cherry picked from commit e93465f) ref #13872

Vectorize @ inbounds for x in A ...

c57eb14

This would previously have been an infinite loop if `length(A) == typemax(Int)` so the loop vectorizer couldn't compute a trip count. (cherry picked from commit fa89a6e) ref #13866

Fix type-instability of eachindex(A, B, C...) (fixes #13870)

d011975

(cherry picked from commit 23c3139) ref #13874

Fix next(::GraphemeIterator) to return SubString

61230f2

This is consistent with eltype() and therefore collect(). Fixes #9261. (cherry picked from commit f379e08) ref #13903

Fix [ind,find][min,max] for LinearFast AbstractArrays (fixes #13888)

5f45e8c

(cherry picked from commit db1ab4a) ref #13909

move sidebars to always be before a section header or end of doc

cc07e6e

(cherry picked from commit 3cf7a34)

fix more jacked up RST all over the place

d651e46

(cherry picked from commit 50c7318)

fix another RST misformat that I missed

dfc1904

(cherry picked from commit f0dbba6)

make RST formatting fix permanent

d9ff642

(cherry picked from commit bbec291)

tkelman mentioned this pull request Nov 7, 2015

Inference failure for typed concatenation #13254

Closed

tkelman added a commit that referenced this pull request Nov 8, 2015

Merge pull request #13834 from JuliaLang/tk/backports-0.4.1

3660d47

[release-0.4, RFC] backports for 0.4.1

tkelman merged commit 3660d47 into release-0.4 Nov 8, 2015

tkelman deleted the tk/backports-0.4.1 branch November 8, 2015 10:25

tkelman mentioned this pull request Nov 30, 2015

[release-0.4, RFC] backports for 0.4.2 #14186

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[release-0.4, RFC] backports for 0.4.1 #13834

[release-0.4, RFC] backports for 0.4.1 #13834

tkelman commented Oct 31, 2015

timholy commented Nov 1, 2015

timholy commented Nov 1, 2015

hiccup7 commented Nov 3, 2015

tkelman commented Nov 4, 2015

tkelman commented Nov 4, 2015

hiccup7 commented Nov 4, 2015

jakebolewski commented Nov 4, 2015

tkelman commented Nov 4, 2015

tkelman commented Nov 7, 2015

[release-0.4, RFC] backports for 0.4.1 #13834

[release-0.4, RFC] backports for 0.4.1 #13834

Conversation

tkelman commented Oct 31, 2015

timholy commented Nov 1, 2015

timholy commented Nov 1, 2015

hiccup7 commented Nov 3, 2015

tkelman commented Nov 4, 2015

tkelman commented Nov 4, 2015

hiccup7 commented Nov 4, 2015

jakebolewski commented Nov 4, 2015

tkelman commented Nov 4, 2015

tkelman commented Nov 7, 2015