[release-0.5] final backports for 0.5.0 #18546

tkelman · 2016-09-16T17:24:57Z

closes #17418

will run pkgeval, and
@nanosoldier runbenchmarks(ALL, vs = ":release-0.5")

…lizer fix #18449 (the real, non-buggy fix is in #18413 for v0.6-dev master)

(cherry picked from commit 788ce67) ref #18514 and #18375

(cherry picked from commit c1ad0cb) ref #18355

Also adds a broken test for #18307 (cherry picked from commit e242d46) ref #18352

(cherry picked from commit e531600) ref #18333

(cherry picked from commit 4e845df) ref #18322

(cherry picked from commit d80fedf) ref #18324

* Add accuracy note for 2-arg log Addresses #18298. * update based on comments (cherry picked from commit 5f74d16)

Fix #18385 (cherry picked from commit c21b22f) ref #18386

Preserve/update `NewvarNode`. (cherry picked from commit 7dc06b0) ref #18386

(cherry picked from commit 92de96b) ref #18392

we want cache=false to disable putting the method in mt->cache (possibly with a wider signature) but it should still be doing the expected unique-ing through the method->specializations cache fix #18404 (cherry picked from commit 9402af3) ref #18415

I am testing julia on a machine where my home directory is a symlink. Unfortunately, this means the repl tests hang, as they wait for the homedir to be given at the shell prompt, but instead the pwd() is printed. `pwd()` calls `uv_cwd()` which calls `getcwd()`, which [will always resolve symbolic links](http://stackoverflow.com/a/1545838/1558890), thus printing something other than `homedir()` on my system. This changes the repl tests to wait for `realpath(homedir())` instead of `homedir()`. (cherry picked from commit f26aee2) ref #18416

(cherry picked from commit 59c1cda) ref #18393

(cherry picked from commit df99e20) ref #18393

also avoid an expensive loop in the common case of the hotpath fix #18399 (cherry picked from commit a188f95) ref #18428

fix #18080 (cherry picked from commit a78ad7b) ref #18430

It is getting installed correctly (cherry picked from commit 5f91c39) ref #18418

Fixes 18395 (cherry picked from commit fa4c02c)

(cherry picked from commit fb503a0)

Has been broken since 172fdb2. (cherry picked from commit 46773af) ref #18536

…iarc Pass through `ENV["PATH"]` in the `setenv` call since it is read from in https://github.com/JuliaLang/julia/blob/v0.5.0-rc4/contrib/mac/juliarc.jl (cherry picked from commit db01a8f) ref #18545 also make the test a little more robust against contents of /tmp (cherry picked from commit ffbb0eb)

musm · 2016-09-16T18:06:42Z

I would add #18368
Since the docs for that are wrong

MichaelHatherly · 2016-09-16T19:24:08Z

#18229 should be fine to backport as well — if it's not too late?

Fixes #11249 to follow the CommonMark spec for lists which does not require a blank line prior to the start of a list. (cherry picked from commit d8da082) ref #18229

nanosoldier · 2016-09-16T20:01:25Z

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @jrevels

tkelman · 2016-09-16T21:43:30Z

binaries for testing
https://julianightlies.s3.amazonaws.com/bin/linux/x64/0.5/julia-0.5.0-b9ec73c89f-linux64.tar.gz
https://julianightlies.s3.amazonaws.com/bin/linux/x86/0.5/julia-0.5.0-b9ec73c89f-linux32.tar.gz
https://julianightlies.s3.amazonaws.com/bin/linux/ppc64le/0.5/julia-0.5.0-b9ec73c89f-linuxppc64.tar.gz
https://julianightlies.s3.amazonaws.com/bin/linux/arm/0.5/julia-0.5.0-b9ec73c89f-linuxarm.tar.gz
https://julianightlies.s3.amazonaws.com/bin/osx/x64/0.5/julia-0.5.0-b9ec73c89f-osx.dmg
https://julianightlies.s3.amazonaws.com/bin/winnt/x64/0.5/julia-0.5.0-b9ec73c89f-win64.exe
https://julianightlies.s3.amazonaws.com/bin/winnt/x86/0.5/julia-0.5.0-b9ec73c89f-win32.exe

~~ARM and Windows ones are still building but should be up before too long.~~

tkelman · 2016-09-17T10:21:17Z

Failures were in Cairo, Mamba, and Gadfly. Cairo we've seen before, might be fixed in that package but need a tag I think? Mamba and Gadfly should be fixed by #18452 which I'll probably merge into this branch but does need a test.

edit: forgot this part https://gist.github.com/106fdba4f38658b0622742f96798a840

this ensures that the writes are completed before flushing the icache otherwise, platforms like ARM would potentially fail to see some of the writes and were especially likely to miss the relocations applied to the JIT code fix #17577 (cherry picked from commit 6f57c40) ref #18516

…on and promotion man (#18368) * Remove uneccesary promote_rule from doc for conversion and promotion * remove promote_op reference since it's dep * Fixed oversights. (cherry picked from commit 7423d72)

Fixes #18531. (cherry picked from commit adcbf1e)

* fix docs for how to change colors in REPL * double backticks (cherry picked from commit 07cf50f)

[release 0.5] hack in a fix to getting the right lambda for the incremental deserializer

Fixes #18271. Allow multidimensional logical arrays as indices into a particular dimension so long as there is only one non-singleton dimension. (cherry picked from commit 5fa5b7f) ref #18401 Fixup inline comment (cherry picked from commit 2bbcd80) Update devdocs for bounds checking following #18401 I had forgotten that this behavior was documented in the developer section of the manual when authoring #18401. (cherry picked from commit b8406ff) ref #18552

tkelman · 2016-09-17T13:09:29Z

@nanosoldier runbenchmarks(ALL, vs = ":release-0.5")

tkelman · 2016-09-17T13:13:32Z

@jrevels recent nanosoldier webhooks aren't going through, "We had a problem connecting to the server. The most common problem with these types of errors is a misconfigured SSL certificate. Try verifying your certificate chain."

* Attempt to fix osx travis We're now getting > Error: You must `brew link autoconf` before staticfloat/julia/julia can be installed Probably because Homebrew officially started supporting 10.12 Sierra, and that broke something. Try to use the 10.10 Travis OSX image instead of the default 10.9 (https://docs.travis-ci.com/user/osx-ci-environment/) * explicitly install cmake on osx travis (cherry picked from commit 45bc17d)

tkelman · 2016-09-18T12:04:34Z

@nanosoldier runbenchmarks(ALL, vs = ":release-0.5")

nanosoldier · 2016-09-18T14:39:50Z

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @jrevels

tkelman · 2016-09-18T17:19:15Z

https://gist.github.com/525cd694db739bc38a7b17761fbaa2d8

same Cairo segfault we've seen before, I started this run before making the tag with the fix
GeoIP didn't finish running properly in the 0.5.0-rc4 baseline run for some reason, but that package is problematic anyway http://pkg.julialang.org/detail/GeoIP.html
NullableArrays looks like the same issue we were seeing in [release-0.4] backports for 0.4.7 #18478 (comment) and Avoid a DomainError if v2 ever happens to be typemin of a signed inte… JuliaStats/NullableArrays.jl#146, probably worked around by the latest tag

Given @timholy's comment at #18401 (comment), I think I'm going to revert 15096f0 (cc @mbauman) and rerun pkgeval one hopefully-last time for good measure.

This reverts commit 15096f0. See the comment at #18401 (comment)

timholy · 2016-09-18T23:11:51Z

Re #18401: out of curiosity, did you get any failures because of it? It is a tough call: on one hand it might conflict with a "better" way to solve the problem for 0.6, on the other hand it fixes a regression from 0.4. I tend to be conservative by not adding behavioral changes at the last minute, but I recognize this is a difficult decision.

tkelman · 2016-09-19T02:00:12Z

From that most recent pkgeval run, it neither caused nor solved any new failures afaict. However note that PyPlot is broken right now so any regressions in packages using it might have been missed.

tkelman · 2016-09-19T03:34:29Z

I'm about to merge and tag this. https://gist.github.com/636a2c609ac0ce380053f0973eb47746
Speak up soon if you think something's missing or I should reconsider any of what's here.

ViralBShah · 2016-09-19T14:41:57Z

Is #17602 release blocking, or should it be tagged 0.5.x?

tkelman · 2016-09-19T14:50:29Z

If there were a PR last week, or a repeatable demonstration of an issue there then maybe, but without a proposed PR I don't think we can hold this for anything else.

vtjnash and others added 22 commits September 11, 2016 23:14

hack in a fix to getting the right lambda for the incremental deseria…

b90f1a2

…lizer fix #18449 (the real, non-buggy fix is in #18413 for v0.6-dev master)

Check for paths instead of path strings

cfbb42f

(cherry picked from commit 788ce67) ref #18514 and #18375

meet_tvars: return bound tvar if both inputs are bound. Fixes #17648

1666d7c

(cherry picked from commit c1ad0cb) ref #18355

Disambiguate sub2ind

9ac7312

Also adds a broken test for #18307 (cherry picked from commit e242d46) ref #18352

Updated Juno documentation shortcut

aa1b840

(cherry picked from commit e531600) ref #18333

Install julia_version.h from BUILDROOT

3b7d29a

(cherry picked from commit 4e845df) ref #18322

Fix dependencies for sysimg out-of-tree builds

8af23c3

(cherry picked from commit d80fedf) ref #18324

Add accuracy note for 2-arg log (#18360)

4a681b7

* Add accuracy note for 2-arg log Addresses #18298. * update based on comments (cherry picked from commit 5f74d16)

Do not swallow UndefVarError in alloc_elim_pass!

21e0998

Fix #18385 (cherry picked from commit c21b22f) ref #18386

Fix for a similar case in alloc_elim_pass!

68ad8d8

Preserve/update `NewvarNode`. (cherry picked from commit 7dc06b0) ref #18386

show replacement for module deprecation

a0fc08d

(cherry picked from commit 92de96b) ref #18392

Fix build_sysimg on windows

6d33adc

(cherry picked from commit 59c1cda) ref #18393

Remove inference0

99df770

(cherry picked from commit df99e20) ref #18393

recompute sparams during inference after widening a method signature

6955092

also avoid an expensive loop in the common case of the hotpath fix #18399 (cherry picked from commit a188f95) ref #18428

fix compile=all

2b4c040

fix #18080 (cherry picked from commit a78ad7b) ref #18430

Remove ppc specific treatment in unwind.mk.

4de501d

It is getting installed correctly (cherry picked from commit 5f91c39) ref #18418

Use typed_hcat in typed_hvcat (#18398)

de37d5d

Fixes 18395 (cherry picked from commit fa4c02c)

fix docstring for print_matrix, fixes #18439

b7be1c0

(cherry picked from commit fb503a0)

Fix compilation with MEMDEBUG2 flag

682a397

Has been broken since 172fdb2. (cherry picked from commit 46773af) ref #18536

tkelman added this to the 0.5.0 milestone Sep 16, 2016

tkelman mentioned this pull request Sep 16, 2016

Simplify and relax logical indexing bounds checks #18401

Merged

Fix doctest line numbers

cf6395f

tkelman mentioned this pull request Sep 16, 2016

Checklist of steps towards 0.5 release [candidates] #17418

Closed

16 tasks

Don't require blank line before markdown lists

b9ec73c

Fixes #11249 to follow the CommonMark spec for lists which does not require a blank line prior to the start of a list. (cherry picked from commit d8da082) ref #18229

tkelman mentioned this pull request Sep 17, 2016

[release 0.5] hack in a fix to getting the right lambda for the incremental deserializer #18452

Merged

vtjnash and others added 6 commits September 17, 2016 04:10

Remove unnecessary/deprecated promote_rule & promote_op from conversi…

d62d212

…on and promotion man (#18368) * Remove uneccesary promote_rule from doc for conversion and promotion * remove promote_op reference since it's dep * Fixed oversights. (cherry picked from commit 7423d72)

Reword doctest description (#18539)

fb96ca7

Fixes #18531. (cherry picked from commit adcbf1e)

fix docs for how to change colors in REPL (#18506)

8ee99a5

* fix docs for how to change colors in REPL * double backticks (cherry picked from commit 07cf50f)

Merge pull request #18452 from JuliaLang/jn/18449

aad3046

[release 0.5] hack in a fix to getting the right lambda for the incremental deserializer

Revert "Simplify and relax logical indexing bounds checks"

feb57e6

This reverts commit 15096f0. See the comment at #18401 (comment)

tkelman mentioned this pull request Sep 19, 2016

non-x86: ensure cgmemmgr caches are consistent #18516

Merged

tkelman merged commit db7eaf9 into release-0.5 Sep 19, 2016

tkelman deleted the tk/backports-0.5.0 branch September 19, 2016 17:19

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[release-0.5] final backports for 0.5.0 #18546

[release-0.5] final backports for 0.5.0 #18546

tkelman commented Sep 16, 2016

musm commented Sep 16, 2016 •

edited

MichaelHatherly commented Sep 16, 2016

nanosoldier commented Sep 16, 2016

tkelman commented Sep 16, 2016 •

edited

tkelman commented Sep 17, 2016 •

edited

tkelman commented Sep 17, 2016

tkelman commented Sep 17, 2016

tkelman commented Sep 18, 2016

nanosoldier commented Sep 18, 2016

tkelman commented Sep 18, 2016

timholy commented Sep 18, 2016

tkelman commented Sep 19, 2016 •

edited

tkelman commented Sep 19, 2016

ViralBShah commented Sep 19, 2016

tkelman commented Sep 19, 2016

[release-0.5] final backports for 0.5.0 #18546

[release-0.5] final backports for 0.5.0 #18546

Conversation

tkelman commented Sep 16, 2016

musm commented Sep 16, 2016 • edited

MichaelHatherly commented Sep 16, 2016

nanosoldier commented Sep 16, 2016

tkelman commented Sep 16, 2016 • edited

tkelman commented Sep 17, 2016 • edited

tkelman commented Sep 17, 2016

tkelman commented Sep 17, 2016

tkelman commented Sep 18, 2016

nanosoldier commented Sep 18, 2016

tkelman commented Sep 18, 2016

timholy commented Sep 18, 2016

tkelman commented Sep 19, 2016 • edited

tkelman commented Sep 19, 2016

ViralBShah commented Sep 19, 2016

tkelman commented Sep 19, 2016

musm commented Sep 16, 2016 •

edited

tkelman commented Sep 16, 2016 •

edited

tkelman commented Sep 17, 2016 •

edited

tkelman commented Sep 19, 2016 •

edited