46f8365 | Valentin Churavy | 07 July 2019, 15:40:29 UTC | Make debuginfo emission configurable | 09 July 2019, 21:38:59 UTC |
13db0dc | Jameson Nash | 09 July 2019, 20:32:41 UTC | partr: fix rare segfault (#32530) The write to 'prio' might be arbitrarily delayed, after some other thread has already popped the task. This commit ensures that all writes to that field are always behind the same lock as the modification to the queue itself. fix #32520 | 09 July 2019, 20:32:41 UTC |
f18808d | Hossein Pourbozorg | 09 July 2019, 08:56:09 UTC | doc: fix typo in sort! docstring (#32533) | 09 July 2019, 08:56:09 UTC |
f552fb2 | Peter F Arndt | 08 July 2019, 18:17:40 UTC | Add methods min, max, minmax taking only single Dates.AbstractTime (#32525) Methods `min`, `max` and `minmax` should also be able to take just one `Dates.AbtractTime` argument. See discussion: https://discourse.julialang.org/t/max-and-min-of-single-dates/26074 | 08 July 2019, 18:17:40 UTC |
33a6b99 | tvrtadx | 08 July 2019, 15:25:49 UTC | rational: `signbit(den)` instead of `sign(den) < 0` (#32523) | 08 July 2019, 15:25:49 UTC |
a25e722 | Jeff Bezanson | 08 July 2019, 10:03:05 UTC | fix #32499, regression in struct expr in assignment rhs (#32512) caused by c6c3d72d1cbddb3d27e0df0e739bb27dd709a413 This is a bit of a hack that just moves things around a bit to get a structure more likely to work. | 08 July 2019, 10:03:05 UTC |
d611e1c | janEbert | 08 July 2019, 06:12:20 UTC | docs: Change `1:size(A, i)` to `axes(A, i)` in at-nloops docs (#32519) | 08 July 2019, 06:12:20 UTC |
c351391 | Keno Fischer | 07 July 2019, 20:13:54 UTC | Yet another LLVM patch (#32506) The previous iteration of this wasn't quite right and I didn't catch it before merging it in julia (or upstream for that matter), so here's the fixup patch. | 07 July 2019, 20:13:54 UTC |
89baf83 | Davide Lasagna | 06 July 2019, 21:10:43 UTC | added *larf and *larfg LAPACK wrappers (#32390) * added LAPACK wrappers for *larf and *larfg routines * add fixes suggested by vtjnash | 06 July 2019, 21:10:43 UTC |
a9ad6c2 | Jeff Bezanson | 06 July 2019, 15:55:01 UTC | fix #32488, type intersection regression (#32509) Issue was caused by 74305bf2c43546bf76843b72ff0d268a3fe16920, so this reverts part of the new logic from there. | 06 July 2019, 15:55:01 UTC |
1272185 | Marco | 06 July 2019, 15:47:48 UTC | fix and tests (#32328) | 06 July 2019, 15:47:48 UTC |
1261a25 | Marco | 06 July 2019, 15:47:22 UTC | make getindex O(1) worst-case (#32329) | 06 July 2019, 15:47:22 UTC |
82f1fae | Mustafa M | 06 July 2019, 15:20:30 UTC | Use tempdir in tempname and update tempdir warning in docstring (#32424) | 06 July 2019, 15:20:30 UTC |
a060344 | Marco | 06 July 2019, 05:54:05 UTC | making Bi/Tri/Sym times Diag output a structured matrix (not sparse) (#31889) * making Bi/Tri/Sym times Diag output a structured matrix (not sparse) * fix test errors/typos * adding diag*bi/tri/sym and removing unused union * add mul!(C,::BiTriSym, ::*Diagonal) * fixing ambiguious methods and adjoint/transpose * space * test error changing `transpose`/`adjoint` to `Transpose`/`Adjoint` This will fail until https://github.com/JuliaLang/julia/pull/31889 is merged. * remove transpose | 06 July 2019, 05:54:05 UTC |
e6379da | Eric Davies | 05 July 2019, 19:41:06 UTC | Export istaskfailed (#32300) * Export istaskfailed * Add docstring for istaskfailed * Add news for istaskfailed | 05 July 2019, 19:41:06 UTC |
b2ac7a3 | Daan Huybrechs | 05 July 2019, 19:08:21 UTC | Fix #30944, ranges with non-IEEEFloat types (#30952) | 05 July 2019, 19:08:21 UTC |
cd79fe6 | chethega | 05 July 2019, 18:38:19 UTC | speed up filter of vector (#31929) | 05 July 2019, 18:38:19 UTC |
95ebe8f | Isaiah Norton | 05 July 2019, 18:33:27 UTC | Document finalizer task-switch restriction (#25141) | 05 July 2019, 18:33:27 UTC |
2350c04 | Jeff Bezanson | 05 July 2019, 16:14:01 UTC | fix #32465, inference of splatting unionall NamedTuple types (#32494) | 05 July 2019, 16:14:01 UTC |
8792eb2 | Jacob Quinn | 05 July 2019, 15:20:48 UTC | Merge pull request #31928 from Moelf/date_name_doc Document Dates.*name() integer method | 05 July 2019, 15:20:48 UTC |
729e87f | Daniel Karrasch | 05 July 2019, 08:51:50 UTC | more performant SparseMatricCSC constructors for structured matrices (#32492) | 05 July 2019, 08:51:49 UTC |
0060e13 | Keno Fischer | 04 July 2019, 21:29:32 UTC | More Makefile definitions for targeting emscripten (#32482) | 04 July 2019, 21:29:32 UTC |
4d68846 | Chris Foster | 04 July 2019, 11:49:58 UTC | Fix tab completion in Logging for functions from CoreLogging (#32473) CoreLogging is an implementation detail, but the fact that Logging imports things from there breaks tab completion for the parts of the public API which aren't re-exported. Fix this by explicitly creating bindings for each imported symbol. | 04 July 2019, 11:49:58 UTC |
3d41f27 | Fredrik Ekre | 04 July 2019, 07:47:08 UTC | Upgrade Pkg to v1.2.0-rc3. (#32487) | 04 July 2019, 07:47:08 UTC |
61a0b04 | ZacNugent | 03 July 2019, 20:58:54 UTC | Update ssair.md (#32489) fix typo | 03 July 2019, 20:58:54 UTC |
a63f2e9 | Keno Fischer | 03 July 2019, 20:54:27 UTC | Merge pull request #32485 from JuliaLang/kf/rettypert Switch jl_rettype_inferred sentintel to jl_nothing not NULL | 03 July 2019, 20:54:27 UTC |
4e1dd9d | Jeff Bezanson | 03 July 2019, 20:43:28 UTC | improve printing of method parameter bounds in candidate list (#32484) | 03 July 2019, 20:43:28 UTC |
38b38cf | Matt Bauman | 03 July 2019, 19:30:35 UTC | Address part of #28866, prevent array operations from dropping 0d containers (#32122) This is a simple workaround for the handful of elementwise operations that are defined on arrays _without_ the need for explicit broadcast but use broadcasting (with an extra shape check) in their implementation. These were the only affected cases I could find. | 03 July 2019, 19:30:34 UTC |
b74ef05 | Jeff Bezanson | 03 July 2019, 16:48:08 UTC | fix #32467, parsing macro on `for` inside `do` inside call (#32476) | 03 July 2019, 16:48:08 UTC |
a17d9ae | Daniel Karrasch | 03 July 2019, 14:11:27 UTC | SparseMatrixCSC constructors for structured matrices - Diagonal, Tridiagonal, etc. (#32466) * move "sparse" code for structured matrices to "SparseMatrixCSC" * spaces between arguments * don't use fallback sparse(::AbstractMatrix) for structured matrices * add tests * fix typo in test | 03 July 2019, 14:11:27 UTC |
c935e13 | Holger Teichgraeber | 03 July 2019, 14:00:32 UTC | Update `-h` flag documentation in getting-started (#32480) | 03 July 2019, 14:00:32 UTC |
28724b1 | Jan Sulmont | 03 July 2019, 13:10:51 UTC | filter specialization for Strings (#32460) (#32478) | 03 July 2019, 13:10:51 UTC |
4b6ab68 | Jan Sulmont | 03 July 2019, 06:45:53 UTC | bug fixed in read_to_buffer (#32457) * bug fixed in read_to_buffer * added test for #32397. * switch back to original case * added torture test for long encoded string randomly laced with spaced | 03 July 2019, 06:45:53 UTC |
07af3d2 | Keno Fischer | 03 July 2019, 04:21:55 UTC | Merge pull request #32481 from JuliaLang/kf/typofix Small typo fix | 03 July 2019, 04:21:55 UTC |
9783c91 | Daniel Karrasch | 03 July 2019, 01:26:01 UTC | specialize matrix multiplication with Diagonals (#32362) * specialize matrix multiplication with Diagonals * use broadcasting everywhere | 03 July 2019, 01:26:01 UTC |
ef89587 | Keno Fischer | 03 July 2019, 00:32:07 UTC | Switch jl_rettype_inferred sentintel to jl_nothing not NULL Using NULL requires treating it as a raw pointer and doing an unsafe operation to recover the object reference. We generally try to avoid NULL sentintels in such situations, using nothing instead. Do the same here. | 03 July 2019, 00:44:59 UTC |
7309c7c | Keno Fischer | 02 July 2019, 22:51:52 UTC | Merge pull request #32396 from JuliaLang/kf/emscriptenllvm Support non-BB LLVM build for Emscripten | 02 July 2019, 22:51:52 UTC |
152030f | Nabanita Dash | 02 July 2019, 22:36:49 UTC | Throw ArgumentError for negative array dimensions and added test (#32363) | 02 July 2019, 22:36:49 UTC |
458b449 | Kristoffer Carlsson | 02 July 2019, 21:55:33 UTC | add an example of programatically creating a named tuple (#32330) | 02 July 2019, 21:55:33 UTC |
b8148ec | Matt Bauman | 02 July 2019, 21:53:42 UTC | Add editor line support and write-only support for rmate (#32357) I often use [rmate](https://github.com/aurora/rmate) to support working on remote systems -- this adds line number support as well as uses the `-f` flag to force it to open read-only files | 02 July 2019, 21:53:41 UTC |
49fc01a | Michael Abbott | 02 July 2019, 21:12:29 UTC | efficient subset query for unit ranges (#32463) | 02 July 2019, 21:12:29 UTC |
2069c2b | Keno Fischer | 02 July 2019, 20:31:30 UTC | Small typo fix Whoops, there's only one underscore here. | 02 July 2019, 20:31:30 UTC |
1fc51b7 | Keno Fischer | 02 July 2019, 20:03:14 UTC | Add simple cross compile mechanism for flisp/libsupport (+ dependencies) (#30526) This makes some incremental progress towards #30338 in the build system. In particular, it allows compiling a separate boostrap version of flisp and its dependencies for an architecture different from the target. The immediate use case here is to build libjulia for wasm. Of course, there's still the enormous problem of building the system image itself, but luckily wasm is largely memory-layout-identical to linux32, so doing that limited bootstrap is fairly easy after this change (though the subject of a different PR). The mechanism here is to use the out-of-tree build support to create separate host/ subdirectories for the build directories of interest. These have a special make variable set that tell it to include a different Make.user (or in the absence thereof use the host defaults). | 02 July 2019, 20:03:14 UTC |
bafb6dd | Keno Fischer | 01 July 2019, 21:11:30 UTC | Carry patch that allows building LLVM with the wasm backend | 02 July 2019, 19:15:20 UTC |
62a9efd | Keno Fischer | 02 July 2019, 19:11:27 UTC | Merge pull request #32471 from JuliaLang/kf/splitproccmd Split Cmd out of process.jl | 02 July 2019, 19:11:27 UTC |
2d810b5 | Keno Fischer | 02 July 2019, 16:53:15 UTC | Merge pull request #32469 from JuliaLang/kf/emscriptenlongjmp Misc backports from kf/wasm3 | 02 July 2019, 16:53:15 UTC |
82883a0 | Keno Fischer | 02 July 2019, 16:52:50 UTC | Fix rounding of Float32->Float16 near eps(Float16(0.0)) (#32444) * Move F16 table to a better place * Fix rounding of Float32->Float16 near eps(Float16(0.0)) The originally described algorithm worked only for round-to-zero, and our adjustment to make it round-to-nearest was incorrect for values that result in Float16 subnormals. Fix this at the expense of an extra `|` and `&`. | 02 July 2019, 16:52:50 UTC |
0eed27e | Valentin Churavy | 02 July 2019, 16:18:00 UTC | Merge pull request #31012 from JuliaLang/jr/checkgeneratedpure fix to allow at-pure optimization for generated thunks that have the pure field set | 02 July 2019, 16:18:00 UTC |
07c2ecc | Jeff Bezanson | 02 July 2019, 13:59:53 UTC | fix the fix for #32121, more named tuple macro hygiene (#32464) | 02 July 2019, 13:59:53 UTC |
a3ac7b9 | Jeff Bezanson | 02 July 2019, 13:59:03 UTC | run threads test with and without `nthreads > 1` (#32284) | 02 July 2019, 13:59:03 UTC |
12b08fc | Keno Fischer | 02 July 2019, 00:27:15 UTC | Split Cmd out of process.jl This change is part of #30534 (libuv disable flag), but I think it makes sense generally, since Cmd and Process are two orthogonal concepts, and I think it makes sense to have one (Cmd, a general representation of a POSIX argument list) and not the other (e.g. in cases where you want to a command to a remote, or more trivially to have the parsing macro available even when you can't spawn a process). | 02 July 2019, 00:27:15 UTC |
710e43d | cfborges | 01 July 2019, 22:09:06 UTC | Correctly rounded variant of the hypot code (#32345) * Implements the correctly rounded variant of the hypot code only in the case where there is a native FMA * Adds comments explaining the functioning of the two branches in the computation. | 01 July 2019, 22:09:06 UTC |
da5a384 | Keno Fischer | 01 July 2019, 22:01:07 UTC | JL_THREAD -> JL_THREAD_LOCAL This code is currently dead on all supported platforms, which is why nobody noticed this breakage, but thread locality is annotated using the `JL_THREAD_LOCAL` macro these days, not the `JL_THREAD` macro (which no longer exists). | 01 July 2019, 22:01:07 UTC |
6589e63 | Keno Fischer | 01 July 2019, 21:57:03 UTC | Fix missing errno header on musl build Backport from kf/wasm3 | 01 July 2019, 21:57:03 UTC |
acacde4 | Keno Fischer | 01 July 2019, 21:23:47 UTC | [wasm] Use regular setjmp/longjmp on Emscripten Emscripten doesn't emulate signals, so sigsetjmp isn't available. Regular setjmp/longjmp works just fine. | 01 July 2019, 21:25:49 UTC |
4911853 | Keno Fischer | 22 June 2019, 23:54:53 UTC | Support non-BB LLVM build for Emscripten When targeting Emscripten/wasm, we were in the odd position that we had BinaryBuilder support, but not in-tree support for building LLVM. Unfortunately, the BinaryBuilder version has recently stopped working due to ABI changes in Emscripten. Use that as an opportunity to backfill the requisite definitions to make an in-tree build of LLVM work on Emscripten. | 01 July 2019, 21:10:22 UTC |
6341e3e | Keno Fischer | 01 July 2019, 20:18:34 UTC | Carry LLVM patch that limits llvm-config flags (#32400) This is an LLVM 6.0/7.0 backport of a patch that changes which flags llvm-config passes. This shouldn't matter much for most uses cases, but it's useful for two reasons: 1) It makes the behavior of llvm-config consistent between all the LLVM versions we support, making it easier for find any problems with our Makefile. 2) It prevents any flags meant for LLVM only from creeping into the build of our .cpp files (there isn't usually a need for this, but I needed to be able to do this when trying to target wasm) | 01 July 2019, 20:18:34 UTC |
98168bc | Keno Fischer | 01 July 2019, 19:29:37 UTC | Merge pull request #32415 from JuliaLang/kf/wasmbuildmisc Two small build fixes under Emscripten | 01 July 2019, 19:29:37 UTC |
df9eb7d | Keno Fischer | 27 June 2019, 22:50:48 UTC | Fix rounding of Float32->Float16 near eps(Float16(0.0)) The originally described algorithm worked only for round-to-zero, and our adjustment to make it round-to-nearest was incorrect for values that result in Float16 subnormals. Fix this at the expense of an extra `|` and `&`. | 01 July 2019, 18:59:44 UTC |
7fa1332 | Keno Fischer | 01 July 2019, 18:58:59 UTC | Move F16 table to a better place | 01 July 2019, 18:58:59 UTC |
00d8694 | Elliot Saba | 01 July 2019, 15:56:09 UTC | base/stat.jl: Mask `UV_EINVAL` error codes for `jl_stat` (#31999) base/stat.jl: Mask `UV_EINVAL` error codes for `jl_stat` | 01 July 2019, 15:56:09 UTC |
99779db | Jarrett Revels | 08 February 2019, 17:29:34 UTC | fix to allow at-pure optimization for generated thunks that have the pure field set Co-authored-by: Valentin Churavy <v.churavy@gmail.com> | 30 June 2019, 03:06:07 UTC |
8d4f6d2 | Valentin Churavy | 28 June 2019, 18:40:26 UTC | Merge pull request #32447 from JuliaLang/jn/llvm8 build fix [NFC] | 28 June 2019, 18:40:26 UTC |
7699238 | Mustafa M | 28 June 2019, 15:00:26 UTC | docs: Add gfortran-mingw-w64 to build instructions (#32445) | 28 June 2019, 15:00:26 UTC |
6e6bd8e | Mustafa M | 28 June 2019, 14:55:14 UTC | Prevent download on windows picking up a rogue curl executable (#32412) | 28 June 2019, 14:55:14 UTC |
032ec82 | Jameson Nash | 27 June 2019, 18:38:59 UTC | compiler: fix LLVM-8 build | 28 June 2019, 14:35:20 UTC |
9d2dfe4 | Jameson Nash | 03 May 2019, 20:18:00 UTC | build: remove duplicate definitions [NFC] | 28 June 2019, 14:35:20 UTC |
8498695 | Viral B. Shah | 28 June 2019, 13:03:59 UTC | Update URL to use new NumFocus URL | 28 June 2019, 13:03:59 UTC |
67cdc55 | Jeff Bezanson | 27 June 2019, 18:41:10 UTC | add thread-safety warnings to global rounding and precision docs (#32426) closes #27139 | 27 June 2019, 18:41:10 UTC |
d2e7d05 | Jeff Bezanson | 27 June 2019, 18:39:50 UTC | fix #16678, add a note about when `:` symbol quoting syntax works (#32436) | 27 June 2019, 18:39:50 UTC |
92f8c50 | Jerry Ling | 27 June 2019, 18:39:23 UTC | \leq and \geq synonyms (#32434) | 27 June 2019, 18:39:23 UTC |
f77743c | Jeff Bezanson | 27 June 2019, 18:29:36 UTC | fix #30653, call `malloc_trim` sometimes (#32428) this works around what seems to be a glibc bug | 27 June 2019, 18:29:36 UTC |
0d9c72d | Jeff Bezanson | 27 June 2019, 18:27:27 UTC | fix #32386, type intersection bug in var bounds (#32425) might be related to #24333 and/or #21153 | 27 June 2019, 18:27:27 UTC |
8ff014f | cossio | 27 June 2019, 12:48:42 UTC | sum for diagonal (and related) matrices (#32184) | 27 June 2019, 12:48:42 UTC |
042504d | Morten Piibeleht | 27 June 2019, 09:59:12 UTC | Add a few methods to manual & fix doctests (#32417) * Don't share doctest state between docstrings This is not, strictly speaking, well defined. There is no official rule as to which docstring should get executed first. This will not be supported in a future version of Documenter. * Add Channel(::Function) to the manual Also fix the doctest in the docstring. The simpler println variant seems to be more reliable when printing from a another task (with at-show, the "Hello" string did not actually get printed). * Fix Base.splat doctest The function is not exported. Note that it does not currently get doctested, but will be tested in a future Documenter version. * Fix doctest in at-locals docstring This is not currently included in the manual and therefore not doctested, but will be in the future. * Fix doctest in Base.isambiguous docstring Not currently doctested since it is not included in the manual, but will be in the future. * Add findnext/last(::AbstractChar,...) to manual Also fix a doctest in one of the docstrings. * Add BLAS.dot and axpby! to manual And fix the doctests in the docstrings. * Fix doctests of unexported functions Fix doctests in the docstrings of LinearAlgebra.isbanded, LinearAlgebra.promote_leaf_eltypes and SparseArrays.dropstored!. All the functions are unexported and not included in the manual. They are currently not being doctested, but will be with a future version of Documenter. | 27 June 2019, 09:59:11 UTC |
7bda2c1 | Morten Piibeleht | 27 June 2019, 09:57:27 UTC | Fix Distributed.head_and_tail (#32431) The last return statement is currently referring to a non-existent s variable. Also fix the related doctest, even though it does not get tested at the moment. | 27 June 2019, 09:57:27 UTC |
fa766bc | chethega | 26 June 2019, 21:36:46 UTC | Speed up filtering of dicts by skipping unneeded hash evaluations (#31242) | 26 June 2019, 21:36:46 UTC |
90fc054 | Jeff Bezanson | 26 June 2019, 21:32:52 UTC | Merge branch 'piever-pv/inet' | 26 June 2019, 21:32:52 UTC |
8453aeb | Pietro Vertechi | 23 March 2019, 11:20:07 UTC | add InetAddr constructor from string (#31459) | 26 June 2019, 21:32:26 UTC |
6f5cefb | Jameson Nash | 26 June 2019, 20:49:25 UTC | arrayunset: micro-optimization for performance (#32405) | 26 June 2019, 20:49:24 UTC |
b32c1b8 | Klaus Crusius | 26 June 2019, 19:41:07 UTC | Argument checks for SparseMatrixCSC constructors (#31724) * reconstruct PR #31118 * reconstruct PR 31118 2 * Check arguments of SparseMatrixCSC #31024 #31435 * fix SuiteSparse test * added NEWS, fixed tests * loosen restrictions - resize to useful length * cleaned up NEWS, revert minor change * add non-checking and checking constructor - improve check performance | 26 June 2019, 19:41:07 UTC |
d6b6cb5 | Alexander Seiler | 26 June 2019, 18:46:26 UTC | Improve performance of opnorm for (m x 1) and (1 x n) sparse matrices (#32372) * Improve performance of opnorm for (m x 1) and (1 x n) sparse matrices * Restrict values of p to 1, 2, Inf * Use nzvalview instead of A.nzval | 26 June 2019, 18:46:26 UTC |
a7427aa | Jameson Nash | 26 June 2019, 18:46:08 UTC | codegen,tbaa: fix array isassigned tbaa information (#32356) This avoids a regression (correctness and performance) caused by #21262 where we were no longer able to fold away the isnull assertion in the case where the source also contains an explicit isassigned check. Also upgrade many other CreateInBoundsGEP calls to include the element type (NFC). | 26 June 2019, 18:46:08 UTC |
7bd700f | Jameson Nash | 26 June 2019, 18:44:40 UTC | lowering,new: improve implementation robustness (#32404) - move implementation of splatnew into lowering instead of dispatch (given the current definition of splatnew) - use field indexes with fieldtype instead of names - use fieldtype in most cases, instead of copying expression tree - explicit syntax error for too many type parameters on new (instead of deferring to the runtime apply_type error) | 26 June 2019, 18:44:40 UTC |
b2304c5 | Jameson Nash | 26 June 2019, 16:23:24 UTC | codegen,gc-lowering: post fixup tbaa information (#32321) Certain metadata (invariant.load, tbaa_const) allows reordering of memory operations over unknown calls (including safepoint). Before GC lowering this is legal, because the optimizer sinking a load over a safepoint will automatically extend the live range of the heap object (by definition the live range of the heap object before GC lowering matches the live ranges of any reference to it). However, after GC lowering, we have fixed the live ranges of the heap objects, and thus it is no longer legal to sink such memory operations over safepoints (as the update to the object's live range won't be tracked). fix #32215 | 26 June 2019, 16:23:24 UTC |
2b08860 | Mustafa M | 26 June 2019, 16:05:47 UTC | Use libuv for tempdir function (#31434) Fixes the Windows tempdir function returning a path with the trailing slash. | 26 June 2019, 16:05:47 UTC |
f7f482b | Daniel Karrasch | 26 June 2019, 15:24:47 UTC | add specific Adjoint{StridedMatrix} * SparseVector method (#32403) add complex and adjoint tests | 26 June 2019, 15:24:47 UTC |
d97d399 | Jeff Bezanson | 26 June 2019, 14:35:36 UTC | speed up `ReentrantLock` by sinking try blocks (#32406) | 26 June 2019, 14:35:36 UTC |
d958419 | Benoît Legat | 26 June 2019, 07:31:35 UTC | Fix element type of -Symmetric(A) for e.g. Bool or JuMP.VariableRef, fixes #32374 (#32375) | 26 June 2019, 07:31:34 UTC |
4671644 | Keno Fischer | 26 June 2019, 00:24:13 UTC | Disable computed goto under emscripten Computed goto is apparently supported using the fastcomp/LLVM -> asm.js -> wasm path, but not the upstream LLVM -> wasm path (at least not at the moment). For now simply always disable computed goto under Emscripten. | 26 June 2019, 00:24:13 UTC |
cb1e583 | Keno Fischer | 26 June 2019, 00:22:17 UTC | Propagate -fPIC flag to LLVM build | 26 June 2019, 00:22:17 UTC |
7bec296 | Thibaut Cuvelier | 25 June 2019, 18:58:15 UTC | Better error message for deprecated scalar-fill indexed assignment, fixes #29266 (#31085) | 25 June 2019, 18:58:15 UTC |
cef655a | Jeff Bezanson | 24 June 2019, 18:12:52 UTC | fix #31521, make regexes thread-safe (#32381) | 24 June 2019, 18:12:52 UTC |
f6049d6 | Alex Arslan | 22 June 2019, 06:53:22 UTC | Add missing test dependency on Random to Statistics (#32383) Statistics uses Random in its tests but doesn't declare it as a test dependency in its Project.toml. | 22 June 2019, 06:53:21 UTC |
77a2c1e | Jameson Nash | 21 June 2019, 20:54:03 UTC | bump libuv to v2-1.29.1-julia (#32280) Co-Authored-By: Elliot Saba <staticfloat@gmail.com> | 21 June 2019, 20:54:03 UTC |
75c10e4 | Jeff Bezanson | 21 June 2019, 17:18:20 UTC | fix bug in thread self-wakeup (#32382) | 21 June 2019, 17:18:20 UTC |
58bafe4 | Jameson Nash | 21 June 2019, 17:17:53 UTC | threads: expand thread-safe region for I/O (#32309) This should hopefully cover most I/O operations which go through libuv to make them thread-safe. There is no scaling here, just one big global lock, so expect worse-than-single-threaded performance if doing I/O on multiple threads (as compared to doing the same work on one thread). The intention is to handle performance improvement incrementally later. It also necessarily redesigns parts of the UDPSocket implementation to properly handle concurrent (single-threaded) usage, as a necessary part of making it handle parallel (thread-safe) usage. | 21 June 2019, 17:17:52 UTC |
1abeeb7 | Elliot Saba | 10 May 2019, 18:34:12 UTC | base/stat.jl: Mask `UV_EINVAL` error codes for `jl_stat` This fixes https://github.com/JuliaLang/julia/issues/26685, where libuv decided to throw `UV_EINVAL` on windows for invalid-filepaths such as `foo??` or `http://google.com`, instead of returning `UV_ENOENT` as they were previously [0]. This causes windows to throw errors on calls such as `isfile("http://google.com")`, whereas other platforms simply return `false`. This change simply ignores `UV_EINVAL` alongside the other "allowed failure" error codes. [0] https://github.com/libuv/libuv/pull/1088 | 20 June 2019, 22:19:49 UTC |
dd56dbf | Kristoffer Carlsson | 20 June 2019, 07:07:27 UTC | remove mention of old instrumenting profiler from docs (#32291) | 20 June 2019, 07:07:27 UTC |
510db13 | Chris Foster | 20 June 2019, 05:01:45 UTC | Fix tryparse for invalid Chars (#32351) This fix adapts the fast path from UInt32(::Char) to the case of digit parsing. | 20 June 2019, 05:01:45 UTC |