7f090db | Jameson Nash | 23 November 2021, 17:24:35 UTC | optimize: attempt to refine inlining costs Try to reflect that each statement has a cost, however tiny, but increase the overall threshold to balance it. | 24 November 2021, 15:45:26 UTC |
3e61ffa | Jameson Nash | 24 November 2021, 05:50:01 UTC | annotate mapreduce file for performance | 24 November 2021, 15:45:26 UTC |
3a5146f | Daniel Karrasch | 23 November 2021, 22:44:12 UTC | Clean-up of blas.jl w/o functional changes (#43193) Add some chkuplo and chkstride1 | 23 November 2021, 22:44:12 UTC |
b9378fd | Jameson Nash | 23 November 2021, 20:53:05 UTC | Revert "Relocate `llvm-version.h` in codegen to avoid a warning" (#43195) * Revert "Relocate `llvm-version.h` in codegen to avoid a warning (#43181)" This reverts commit a0c920c1d977c177d13d9a328fde8f90f506778a. * remove old define from llvm-version Co-authored-by: Valentin Churavy <v.churavy@gmail.com> | 23 November 2021, 20:53:05 UTC |
99f6558 | Fredrik Ekre | 23 November 2021, 18:04:13 UTC | Fix shell syntax in version_git.sh and remove warning. (#43196) | 23 November 2021, 18:04:13 UTC |
08ea2d8 | Rafael Fourquet | 23 November 2021, 15:41:12 UTC | `@testset`: with Xoshiro, restore Random.GLOBAL_SEED (#43188) A `@testset` is supposed to restore the "global RNG state" as it was before execution (so that they can be re-ordered easily, etc.) Also, before a testset starts, the default RNG is re-seeded with the "global seed" (to help reproduce test failures). Before `Xoshiro` as the default RNG, the "global seed" was stored within a `MersenneTwister` object. It was enough for a testset to copy the default RNG at the start, and copy it back at the end. But now the global seed is stored outside of the RNG, so it should also be restored separately. | 23 November 2021, 15:41:12 UTC |
8c97995 | Jameson Nash | 23 November 2021, 10:15:10 UTC | fix gc collection interval logic (#43136) We had previously prevented full collections for the first collection, forgotten to reset counters for gc_sweep_always_full, and caused resets of other counter heuristics even when not heuristically driven. | 23 November 2021, 10:15:10 UTC |
04d9229 | Jameson Nash | 23 November 2021, 10:14:48 UTC | mark these methods as eligible for the faster cache for novel lookups (#36765) | 23 November 2021, 10:14:48 UTC |
f8defa4 | Alex Arslan | 23 November 2021, 03:51:12 UTC | Fix `-Wnon-c-typedef-for-linkage` from `jl_codegen_params_t` (#43187) Currently `jl_codegen_params_t` is defined as `typedef struct {...}` without a tag name. If we give it an underscore-prefixed tag name (as is the convention used in our other type declarations), Clang will stop complaining about having to assume what the name should be for linkage. | 23 November 2021, 03:51:12 UTC |
3f3632c | Ian Atol | 23 November 2021, 03:35:20 UTC | Relax constraints on inlining for some single calls (#43113) Co-authored-by: Shuhei Kadowaki <aviatesk@gmail.com> | 23 November 2021, 03:35:20 UTC |
9ab57e8 | Kenneth Johnsen | 22 November 2021, 22:37:08 UTC | doc/strings: use string(a, b) instead of * (#42905) The * operator for concatenation is introduces below, so it's less confusing if `string(a,b)` is used here. Also update an outdated repl-block to a jldoctest. | 22 November 2021, 22:37:08 UTC |
eae29c7 | A1-Mathematik | 22 November 2021, 20:37:00 UTC | Fix typo: RecursiveLock to ReentrantLock (#43192) | 22 November 2021, 20:37:00 UTC |
d44a534 | Jameson Nash | 22 November 2021, 18:36:59 UTC | Better handling for Union-type fields, particularly of singletons (#43163) fix #43123 | 22 November 2021, 18:36:59 UTC |
ae336ab | Ben Arthur | 22 November 2021, 12:17:00 UTC | add spr! to LinearAlgebra.BLAS (#42830) | 22 November 2021, 12:17:00 UTC |
a21cc80 | Dilum Aluthge | 22 November 2021, 08:14:13 UTC | `Base.GIT_VERSION_INFO`: record the Buildkite commit that was used to build Julia (#43073) | 22 November 2021, 08:14:13 UTC |
a0c920c | Alex Arslan | 22 November 2021, 06:11:15 UTC | Relocate `llvm-version.h` in codegen to avoid a warning (#43181) Our `llvm-version.h` defines `LLVM_ENABLE_STATS` if it's not already defined. However, we're `#include`ing that file in `codegen.cpp` before we `#include` the LLVM headers that define it. That means it gets defined twice, triggering a `-Wmacro-redefined` warning. If we move the inclusion of our file to below the inclusion of LLVM's headers, we just use their definition and avoid a warning. | 22 November 2021, 06:11:15 UTC |
b078d77 | Nicholas Bauer | 22 November 2021, 02:28:44 UTC | Fix UndefVarError due to unintentional signature change (#43173) * Fix UndefVarError due to unintentional signature change * Add wait for REPL task to exit * Add test * test fix * no invoke | 22 November 2021, 02:28:44 UTC |
d87c384 | DilumAluthgeBot | 21 November 2021, 15:31:28 UTC | 🤖 Bump the Pkg stdlib from b963d05a to 0d6a726e (#43180) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 21 November 2021, 15:31:28 UTC |
a4f7f2d | Marco | 21 November 2021, 04:42:38 UTC | Avoid allocating diagonal zeros in `sparse(::Diagonal)` (#42577) Co-authored-by: Daniel Karrasch <daniel.karrasch@posteo.de> | 21 November 2021, 04:42:38 UTC |
e9430c9 | Dilum Aluthge | 21 November 2021, 01:55:04 UTC | `Base.runtests`: rename the `--force-net` option to `--ci` (#43168) Co-authored-by: Rafael Fourquet <fourquet.rafael@gmail.com> Co-authored-by: Rafael Fourquet <fourquet.rafael@gmail.com> | 21 November 2021, 01:55:04 UTC |
06d5308 | Alex Arslan | 20 November 2021, 21:34:33 UTC | Avoid `-Wunused-function` on macOS for some `Float16` intrinsics (#43174) We're defining `half_to_float` and `float_to_half` unconditionally but only ever calling them on non-Mac platforms, so Clang produces unused function warnings during the build on macOS. To fix this, we can just move the platform check to encompass the definitions in addition to the uses. | 20 November 2021, 21:34:33 UTC |
c2e9311 | Lilith Orion Hafner | 20 November 2021, 19:41:26 UTC | Restrict fpsort! to more homogenous collections (#42740) fixes #42739 | 20 November 2021, 19:41:26 UTC |
1799ae6 | Shuhei Kadowaki | 20 November 2021, 06:51:57 UTC | optimizer: some cosmetic improvements on inlining code (#43115) - remove an unnecessary field and allocation - make the signature order consistent across the file: `ir::IRCode, idx::Int, stmt, rt, info, flag::UInt8, sig::Signature, state::InliningState, todo::Vector{Pair{Int, Any}}` - pass the same `stmt::Expr` object and slightly improve its type-stability | 20 November 2021, 06:51:57 UTC |
2a1197a | Takafumi Arakaki | 20 November 2021, 05:25:52 UTC | Remove stale `cglobal(:jl_tls_states)` test (#43161) * Remove stale `cglobal(:jl_tls_states)` test * Remove `jl_threading_enabled` * The function `threadingOn()` in `julia-config.jl` is the only reference other than the tests just removed. It looks like `threadingOn` itself is not used anywhere in the script. | 20 November 2021, 05:25:52 UTC |
feba260 | Oscar Smith | 20 November 2021, 04:47:48 UTC | fix bug in fma_emulated (#43156) Co-authored-by: oscarddssmith <oscar.smith@juliacomputing.com> | 20 November 2021, 04:47:48 UTC |
b64f1cd | Mason Protter | 20 November 2021, 00:49:23 UTC | Make `iterate(::Reverse)` use indexing by default (#43110) Co-authored-by: Jeff Bezanson <jeff.bezanson@gmail.com> Co-authored-by: Jeff Bezanson <jeff.bezanson@gmail.com> Co-authored-by: Jameson Nash <vtjnash@gmail.com> | 20 November 2021, 00:49:23 UTC |
f9e40ef | Michiel Dral | 20 November 2021, 00:29:15 UTC | Allow `X.@x() do; end` syntax, fixes #43018 (#43154) | 20 November 2021, 00:29:15 UTC |
50d1d0f | Shuhei Kadowaki | 20 November 2021, 00:02:51 UTC | optimizer: slightly improve callsite annotation handling (#43131) | 20 November 2021, 00:02:51 UTC |
22a192c | Jeff Bezanson | 19 November 2021, 20:50:54 UTC | speed up resolve-scopes lowering pass (#43022) Use a table to avoid each symbol lookup being O(n) in the scope nesting depth. | 19 November 2021, 20:50:54 UTC |
2c2e724 | Nicholas Bauer | 19 November 2021, 19:26:25 UTC | Add `err` global to REPL to store most recent errors (#40642) | 19 November 2021, 19:26:25 UTC |
d730703 | Éric Thiébaut | 19 November 2021, 19:23:16 UTC | call `reverse(o::Base.Ordering)` to reverse ordering specified by `o` (#42685) | 19 November 2021, 19:23:16 UTC |
7afb3b4 | Tamas K. Papp | 19 November 2021, 19:16:57 UTC | Fix documentation for searchsortedfirst/last w/ generalized indexing. (#43128) While the code now correctly handles generalized indexing, documentation for `searchsortedfirst` and `searchsortedlast` still assume that indices are `1:length(a)`. This PR fixes that. Also add tests while at it. No behavior is changed. | 19 November 2021, 19:16:57 UTC |
f5e0f9d | Sacha Verweij | 19 November 2021, 19:15:07 UTC | Make stale_cachefile compatible with Nix mtime (#43090) The added condition improves compatiblity with Nix mtime. | 19 November 2021, 19:15:07 UTC |
2c14d93 | Shuhei Kadowaki | 19 November 2021, 16:37:25 UTC | inference: use `ssa_def_slot` for `InterConditional` propagation (#43147) This tweak should allow constraint propagation in following kind of cases: ```julia julia> isaT(x, T) = isa(x, T); \# backward propagation julia> @test Base.return_types((Any,Int)) do a, b c = a if isaT(c, typeof(b)) return c # c::Int end return 0 end |> only === Int; \# forward propagation julia> @test Base.return_types((Any,Vector{Any})) do x, y z = x ifelse(isa(z, Int), z, length(y)) end |> only === Int; ``` | 19 November 2021, 16:37:25 UTC |
672c641 | Takafumi Arakaki | 19 November 2021, 15:44:51 UTC | Disable "Test spawning a lot of tasks" on Windows (#43125) * Don't hide output from spawnmany test script * Disable "Test spawning a lot of tasks" on Windows * If `!success(proc) || timeout`, print a variety of debugging data Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 19 November 2021, 15:44:51 UTC |
3f8493c | Jeff Bezanson | 19 November 2021, 15:32:02 UTC | make sure `reverse(::Pairs)` works (#43143) fixes #43045 | 19 November 2021, 15:32:02 UTC |
9690237 | Shuhei Kadowaki | 19 November 2021, 06:04:08 UTC | slightly improve base/compiler type-stabilities (#43132) | 19 November 2021, 06:04:08 UTC |
ee21d7b | woclass | 19 November 2021, 06:03:38 UTC | [devdoc] update section "reflection" (#43145) * [devdoc] fix typo + `isbitstype` accepts types, e.g.: `Int64` + `isbits` accepts instances, e.g.: `1` * [devdoc] update REPL output * [devdoc] mark code block type * [devdoc] add space for inline code | 19 November 2021, 06:03:38 UTC |
61c6f33 | Dilum Aluthge | 19 November 2021, 03:44:35 UTC | `contrib/refresh_checksums.mk`: automatically delete old checksums before adding new checksums (#43141) Co-authored-by: Jameson Nash <vtjnash@gmail.com> Co-authored-by: Jameson Nash <vtjnash@gmail.com> | 19 November 2021, 03:44:35 UTC |
2c057f8 | Simeon Schaub | 19 November 2021, 03:37:39 UTC | improve constraint propagation with `&&` in `||` (#43142) I don't know why we previously emitted an SSA value here. We only accept bools in if-conditions, so I don't see how `g` could be anything but `true` inside the first clause. fixes #43130 | 19 November 2021, 03:37:39 UTC |
dde1756 | Christine Flood | 19 November 2021, 00:27:11 UTC | Fix JuliaLang#42774 (#43122) ``` using Test ``` Before: ``` julia> @test :(1 := 2) |> repr == ":($(Expr(:(:=), 1, 2)))" Test Passed Expression: :($(Expr(:(:=), 1, 2))) |> repr == ":($(Expr(:(:=), 1, 2)))" Evaluated: ":(\$(Expr(:(:=), 1, 2)))" == ":(\$(Expr(:(:=), 1, 2)))" ``` After: ``` julia> @test :(1 := 2) |> repr == ":($(Expr(:(:=), 1, 2)))" Test Passed Expression: :(1 := 2) |> repr == ":($(Expr(:(:=), 1, 2)))" Evaluated: ":(1 := 2)" == ":(1 := 2)" ``` Co-authored-by: Christine H. Flood <christineflood@juliacomputing.com> | 19 November 2021, 00:27:11 UTC |
0070f82 | Jameson Nash | 18 November 2021, 21:03:24 UTC | Merge pull request #43106 from JuliaLang/jn/revert-42991 Revert "make last work on any reversible collection (#42991), and fix it | 18 November 2021, 21:03:24 UTC |
b1f50af | Mosè Giordano | 18 November 2021, 20:34:27 UTC | Add space after dot in `checkfor_mv_cp_cptree` error msg (#43134) At the moment the error message reads ``` ERROR: ArgumentError: 'src' and 'dst' refer to the same file/dir.This is not supported. ``` | 18 November 2021, 20:34:27 UTC |
ebcce3f | Bogumił Kamiński | 18 November 2021, 20:03:36 UTC | Improve count performance (#40564) | 18 November 2021, 20:03:36 UTC |
85f4db2 | kcajf | 18 November 2021, 19:42:06 UTC | add chopprefix, chopsuffix (#40995) Co-authored-by: Jameson Nash <vtjnash@gmail.com> | 18 November 2021, 19:42:06 UTC |
616de75 | Rohan Mallick | 18 November 2021, 19:34:24 UTC | Divrem speedup and tests (#39045) fixes #22219 | 18 November 2021, 19:34:24 UTC |
edea1de | Milan Bouchet-Valat | 18 November 2021, 15:04:51 UTC | Fix regression in `map` and `collect` (#43120) `promote_typejoin_tuple` returned a type which was more precise than the one which `map` and `collect` actually use via `promote_type`, triggering an assertion error in corner cases. | 18 November 2021, 15:04:51 UTC |
455236e | Takafumi Arakaki | 18 November 2021, 02:18:01 UTC | Fix stack pointer retrieval in jl_backtrace_from_here (#42585) Co-authored-by: Jameson Nash <vtjnash@gmail.com> | 18 November 2021, 02:18:01 UTC |
0ca6948 | EvianZhang | 18 November 2021, 00:31:26 UTC | Use `open` to open Julia REPL in macOS default terminal (#43111) Most macOS developers use third-party terminal apps (e.g. iTerm2) to replace macOS's default terminal. Current implementation will always open built-in terminal to execute Julia REPL, which is not good enough. | 18 November 2021, 00:31:26 UTC |
b994e0b | Jameson Nash | 18 November 2021, 00:03:08 UTC | make our locks fast (#42964) Now that we have the ability to express atomics (particularly the write barrier of the Task object), it is time for us to speed up and simplify our locks. This PR does need someone to verify the synchronizations are correct, but I think this version is actually quite simple. There is a single byte that represents the state of the lock: 0x0 : unlocked 0x1 : locked (without other waiters) 0x2 : locked with waiters And then when we have the locked bit set, we can manipulate any of the other data (e.g. owner and reentrancy) without needing any atomic work, and if we don't hold that bit, we should not write to any other field. | 18 November 2021, 00:03:08 UTC |
5e4773b | Jeff Bezanson | 17 November 2021, 22:18:55 UTC | speed up `analyze-variables` lowering pass by adding a hash table of variables (#43028) | 17 November 2021, 22:18:55 UTC |
40ffd6f | woclass | 17 November 2021, 21:44:38 UTC | [doc] use autogenerated names, keep existing links usable (#43027) | 17 November 2021, 21:44:38 UTC |
8d5efbb | woclass | 17 November 2021, 21:29:29 UTC | fix generic `findall(A)` (#43089) Fixes #43078; This problem was introduced by #36965 | 17 November 2021, 21:29:29 UTC |
c5b5a76 | Shuhei Kadowaki | 17 November 2021, 15:26:49 UTC | inference: emit special constant-prop' remark for `LimitedAccuracy` rt (#43114) xref: <https://github.com/JuliaDebug/Cthulhu.jl/issues/241#issuecomment-971207306> | 17 November 2021, 15:26:49 UTC |
8dc1fa4 | Katharine Hyatt | 17 November 2021, 14:01:02 UTC | Merge pull request #43024 from JuliaLang/ksh/chartests A few tests for Char and transcode | 17 November 2021, 14:01:02 UTC |
9bd6dbc | Michael Abbott | 17 November 2021, 12:29:21 UTC | Improve docstrings for `union` & `intersect` (#42695) * docstring for union * infix * intersect too * tweak * apply suggestions | 17 November 2021, 12:29:21 UTC |
764d876 | Shuhei Kadowaki | 17 November 2021, 08:19:52 UTC | compiler: minor refactors (#43105) - factor the `typeintersect(A, B) === Bottom` pattern into `hasintersect` - remove `uniontypes` allocation within `isdefined_tfunc` - don't use `==` for comparison to `Bottom` | 17 November 2021, 08:19:52 UTC |
89c137c | Sebastian Pfitzner | 17 November 2021, 05:04:16 UTC | Fix typo in iterate(itr::AsyncCollector) (#43041) | 17 November 2021, 05:04:16 UTC |
65a6bca | Oscar Smith | 17 November 2021, 01:50:44 UTC | Clarify AbstractRNG documentation (#41875) * Clarify AbstractRNG documentation clean up the section a bit. * make the exported types a bulleted list * point out that `MersenneTwister` shouldn't be used except for backwards compatability * explicitly mention `DeviceRandom` is what to use as a secure random generator. * Update stdlib/Random/docs/src/index.md Co-authored-by: Rafael Fourquet <fourquet.rafael@gmail.com> * Update stdlib/Random/docs/src/index.md Co-authored-by: Jameson Nash <vtjnash@gmail.com> * Update stdlib/Random/docs/src/index.md Co-authored-by: Jameson Nash <vtjnash@gmail.com> * Update stdlib/Random/docs/src/index.md Co-authored-by: Jameson Nash <vtjnash@gmail.com> * Update stdlib/Random/docs/src/index.md Co-authored-by: Rafael Fourquet <fourquet.rafael@gmail.com> * Update stdlib/Random/docs/src/index.md Co-authored-by: Jameson Nash <vtjnash@gmail.com> * Update stdlib/Random/docs/src/index.md Co-authored-by: Jameson Nash <vtjnash@gmail.com> Co-authored-by: Rafael Fourquet <fourquet.rafael@gmail.com> Co-authored-by: Jameson Nash <vtjnash@gmail.com> Co-authored-by: Viral B. Shah <ViralBShah@users.noreply.github.com> | 17 November 2021, 01:50:44 UTC |
7fcc854 | Jonas Eschle | 17 November 2021, 00:38:52 UTC | WIP: Clarify performance penalty of global (#42452) * Clarify performance penalty of global I've slightly changed the wording for why a global creates a penalty in performance, as it raised confusion [such as here](https://discourse.julialang.org/t/global-variables-const-by-default/33138/3?) and I think the current wording was wrong (changing the value does not need to change the type, correct?) However, I am still not quite convinced as a variable *doesn't* has a type and therefore "it's type changes" seems a bit weird, doesn't it? * Update performance-tips.md * Update performance-tips.md * Update performance-tips.md * Update performance-tips.md * Update performance-tips.md * Update doc/src/manual/performance-tips.md Co-authored-by: Viral B. Shah <ViralBShah@users.noreply.github.com> Co-authored-by: Fredrik Ekre <ekrefredrik@gmail.com> | 17 November 2021, 00:38:52 UTC |
0722f3c | Simon Byrne | 17 November 2021, 00:29:11 UTC | Document that cfunctions should not throw errors (#43100) * Document that cfunctions should not throw errors This was not documented, see https://discourse.julialang.org/t/what-happens-when-you-throw-an-error-from-a-cfunction/71510?u=simonbyrne * Update doc/src/manual/calling-c-and-fortran-code.md Co-authored-by: Jameson Nash <vtjnash@gmail.com> Co-authored-by: Jameson Nash <vtjnash@gmail.com> | 17 November 2021, 00:29:11 UTC |
df2abc6 | Glen Hertz | 17 November 2021, 00:27:36 UTC | Improve `cis` and `cispi` docs (#40859) * Improve `cis` and `cispi` docs * Fix whitespace error * Fix escaping of \\pi * Add suggestions * Add sincos refs (more discoverable) * Typo fix * Tweaks * More wordsmithing * Fix whitespace at EOL * wordsmithing * More wordsmithing * More wordsmithing * Use larger value of cispi(x) to show accuracy in example | 17 November 2021, 00:27:36 UTC |
a05bcb2 | Chris Foster | 16 November 2021, 23:32:10 UTC | Fix world age issue with custom streams for Distributed workers (#42481) If connect(::CustomClusterManager, ...) returns a custom transport stream, use of that stream by the task in start_gc_msgs_task() may fail due to the task executing in an old world age. Add an invokelatest() to prevent this problem. | 16 November 2021, 23:32:10 UTC |
6202475 | Rafael Fourquet | 16 November 2021, 19:59:11 UTC | add trunc(BigInt, ::Float16) (#40849) | 16 November 2021, 19:59:11 UTC |
5cb1d68 | Katharine Hyatt | 10 November 2021, 13:01:48 UTC | A few tests for Char and transcode | 16 November 2021, 19:29:57 UTC |
5754473 | bzinberg | 16 November 2021, 19:26:21 UTC | Document `isdone` as part of the iterator interface (#43099) It's a necessary method to implement when defining a mutable custom iterator. See https://discourse.julialang.org/t/isempty-enumerate-iterators-stateful-advances-the-stateful-iterator/71475 | 16 November 2021, 19:26:21 UTC |
3049893 | Jameson Nash | 16 November 2021, 16:17:15 UTC | add back `last` for every type that defines `Iterators.reverse` Fixes #42943 Fixes #43101 Replaces #42991 | 16 November 2021, 18:51:18 UTC |
57fa0ca | Jameson Nash | 16 November 2021, 15:49:51 UTC | Revert "make `last` work on any reversible collection (#42991)" This reverts commit 1f484c3ecabdd7519fa68569b1688a153e92fb16 since it seems to cause excess breakage for the benefits it brings. | 16 November 2021, 15:49:51 UTC |
bc8337a | N5N3 | 16 November 2021, 14:47:14 UTC | Fix `==`/`sum`/`issymmetric` for (Sym)Tridiagonal with non-number eltype (#43066) | 16 November 2021, 14:47:14 UTC |
9affe2f | Daniel Karrasch | 16 November 2021, 14:46:41 UTC | Make (Sym)Tridiag-Diagonal solves return TriDiagonal (#42744) | 16 November 2021, 14:46:41 UTC |
73f46aa | woclass | 16 November 2021, 06:40:51 UTC | [libjulia-codegen] export `jl_jit_total_bytes` (#43086) | 16 November 2021, 06:40:51 UTC |
8a8530f | Michele Zaffalon | 16 November 2021, 06:40:16 UTC | Show use of let without immediate assignment (#43094) Co-authored-by: Steven G. Johnson <stevenj@mit.edu> Co-authored-by: Jameson Nash <vtjnash@gmail.com> Co-authored-by: Dilum Aluthge <dilum@aluthge.com> Co-authored-by: Steven G. Johnson <stevenj@mit.edu> | 16 November 2021, 06:40:16 UTC |
d958c8c | Sacha Verweij | 16 November 2021, 03:56:09 UTC | Unbreak source distribution tarball construction. (#43096) Co-authored-by: Jameson Nash <vtjnash@gmail.com> | 16 November 2021, 03:56:09 UTC |
28c49ce | Jameson Nash | 16 November 2021, 02:03:13 UTC | fix atomic fences (#43093) Previously, we were disallowing the opposite of the orders that we intended to disallow. | 16 November 2021, 02:03:13 UTC |
4b99bc8 | Takafumi Arakaki | 15 November 2021, 20:37:15 UTC | Enable "Test spawning a lot of tasks" on non-Linux (#43072) | 15 November 2021, 20:37:15 UTC |
8eb94c6 | Jeff Bezanson | 15 November 2021, 19:28:56 UTC | fix #41727, field doc string with interpolation (#43061) | 15 November 2021, 19:28:56 UTC |
4378572 | Shuhei Kadowaki | 15 November 2021, 19:05:03 UTC | add some unnecessary MethodError restrictions for Type (#43087) The previous `@nospecialize(::Type)` dispatch signatures actually caused the following kinds of dynamic dispatches: ```julia julia> using JET julia> report_opt((Vector{Any},)) do ts isbitstype(ts[1]) end ═════ 1 possible error found ═════ ┌ @ REPL[3]:2 Main.isbitstype(%1) │ runtime dispatch detected: Main.isbitstype(%1::Any) └───────────── ``` This commit also adds an overload of `emptymutable(::IdSet)`, which eliminates the dynamic dispatch below: ```julia julia> report_opt((Base.IdSet{Any}, Base.IdSet{Any}); annotate_types=true) do s1, s2 s1 ∪ s2 end ═════ 2 possible errors found ═════ ┌ @ REPL[3]:2 Main.∪(s1::Base.IdSet{Any}, s2::Base.IdSet{Any}) │┌ @ abstractset.jl:48 Base.union!(Core.tuple(Base.emptymutable(s::Base.IdSet{Any}, Base.promote_eltype(Core.tuple(s::Base.IdSet{Any})::Tuple{Base.IdSet{Any}}, sets::Tuple{Base.IdSet{Any}}...)::Type{Any})::Set{Any}, s::Base.IdSet{Any})::Tuple{Set{Any}, Base.IdSet{Any}}, sets::Tuple{Base.IdSet{Any}}...) ││┌ @ abstractset.jl:76 Base.union!(s::Set{Any}, x::Base.IdSet{Any}) │││┌ @ abstractset.jl:93 Base.sizehint!(s::Set{Any}, Base.+(Base.length(s::Set{Any})::Int64, Core.typeassert(Base.Int(Base.length(itr::Base.IdSet{Any})::Int64)::Int64, Base.Int)::Int64)::Int64) ││││┌ @ set.jl:84 Base.sizehint!(Base.getproperty(s::Set{Any}, :dict::Symbol)::Dict{Any, Nothing}, newsz::Int64) │││││┌ @ dict.jl:244 Base.rehash!(d::Dict{Any, Nothing}, newsz::Int64) ││││││┌ @ dict.jl:203 Base.hashindex(%231::Any, %20::Int64) │││││││ runtime dispatch detected: Base.hashindex(%231::Any, %20::Int64) ││││││└─────────────── │││┌ @ abstractset.jl:95 Base.push!(s::Set{Any}, %62::Any) ││││ runtime dispatch detected: Base.push!(s::Set{Any}, %62::Any) │││└───────────────────── ``` | 15 November 2021, 19:05:03 UTC |
0a4338b | Viral B. Shah | 15 November 2021, 11:56:43 UTC | Update README.md (#43081) The main change here is to recommend only VS Code as the IDE. | 15 November 2021, 11:56:43 UTC |
8131580 | Takafumi Arakaki | 14 November 2021, 01:12:24 UTC | Make profiling more robust with many tasks (#42978) This patch includes two sets of changes. (1) `jl_thread_suspend_and_get_state` uses `pthread_cond_timedwait` to recover from the case where the request is not received by the signal handler. This is required because `usr2_handler` contains some paths for the case where it is not possible to obtain `ptls`. (2) `ctx_switch` now makes sure to null out `ptls` of the last task (`lastt->ptls = NULL`) after changing the current task by updating pgcstack (`jl_set_pgcstack(&t->gcstack)`). This closes the gap in which `usr2_handler` can observe the null `ptls`. Co-authored-by: Jameson Nash <vtjnash@gmail.com> | 14 November 2021, 01:12:24 UTC |
5665e8b | Takafumi Arakaki | 14 November 2021, 01:02:40 UTC | Better handling of racy `@spawn` in `@sync` (#41927) Co-authored-by: Jameson Nash <vtjnash@gmail.com> | 14 November 2021, 01:02:40 UTC |
ddd0e53 | Bogumił Kamiński | 13 November 2021, 18:04:56 UTC | Add iterate for Reverse{<:NamedTuple} (#43038) Fixes the problem mentioned in https://github.com/JuliaLang/julia/pull/42991#issuecomment-966190513 | 13 November 2021, 18:04:56 UTC |
81e3889 | Sacha Verweij | 13 November 2021, 17:36:37 UTC | Merge pull request #43003 from JuliaLang/sv-keep-hidden Stop stripping git, github, and CI config files from source distributions. | 13 November 2021, 17:36:37 UTC |
998960e | DilumAluthgeBot | 13 November 2021, 07:06:01 UTC | 🤖 Bump the Pkg stdlib from 834a6c44 to b963d05a (#43062) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 13 November 2021, 07:06:01 UTC |
68020b0 | Jeff Bezanson | 13 November 2021, 01:27:22 UTC | fix #43040, crash when a concrete method sig has an unused static parameter (#43048) | 13 November 2021, 01:27:22 UTC |
780ebea | Sacha Verweij | 09 November 2021, 00:06:29 UTC | Stop stripping git, github, and CI config files from source distributions. | 12 November 2021, 21:31:43 UTC |
5e2894b | Harmen Stoppels | 12 November 2021, 18:03:50 UTC | Symlink 7z when USE_SYSTEM_P7ZIP (#43005) * Symlink 7z * Simplify Co-authored-by: Elliot Saba <staticfloat@gmail.com> Co-authored-by: Elliot Saba <staticfloat@gmail.com> | 12 November 2021, 18:03:50 UTC |
8cf89d7 | Gustavo Goretkin | 12 November 2021, 17:03:32 UTC | Reduce code duplication in find* functions (#36965) | 12 November 2021, 17:03:32 UTC |
3ef1f61 | Jishnu Bhattacharya | 12 November 2021, 16:24:08 UTC | Define `promote_rule` for `Diagonal` matrices (#42142) Co-authored-by: Daniel Karrasch <daniel.karrasch@posteo.de> | 12 November 2021, 16:24:08 UTC |
17f05b5 | Lilith Orion Hafner | 12 November 2021, 15:43:44 UTC | [NFC] rename where -> counts (#43052) Because `where` is a keyword that often has syntax highlighting, and `counts` is more descriptive. | 12 November 2021, 15:43:44 UTC |
ec3ec02 | Jameson Nash | 12 November 2021, 15:43:02 UTC | codegen: add missing initialization for PhiC nodes (#43029) Our Phi handling assumes that it can references undefined memory, and get back legal results, but our PhiC nodes were not initialized, so the Phi node might see uninitialized results, and then cause the GC to crash. This was observed in PkgEval on the PoreMatMod.jl package to occur in recent Julia versions and master. | 12 November 2021, 15:43:02 UTC |
b71330d | Jameson Nash | 12 November 2021, 15:42:42 UTC | ranges: fix empty length for smallints (#43042) Fixes #29801 | 12 November 2021, 15:42:42 UTC |
dc26322 | Jameson Nash | 12 November 2021, 15:41:57 UTC | fix overflow for empty non-integer-type ranges (#43043) Fixes #29810 | 12 November 2021, 15:41:57 UTC |
d08b05d | Viral B. Shah | 11 November 2021, 23:48:11 UTC | Revert "Update suitesparse url (#43046)" (#43049) This reverts commit 6779a9c98768d21f703ddbc286e27ac06c698ac8. | 11 November 2021, 23:48:11 UTC |
4846ed7 | Jameson Nash | 11 November 2021, 23:08:55 UTC | Merge pull request #43031 from JuliaLang/jn/ast-locks flisp: remove support for and need for reentrancy | 11 November 2021, 23:08:55 UTC |
2cfebad | Jeff Bezanson | 11 November 2021, 23:03:23 UTC | disallow unbalanced bidirectional formatting in strings and comments (#42918) | 11 November 2021, 23:03:23 UTC |
f4f92c5 | Jeff Bezanson | 11 November 2021, 21:40:15 UTC | disable extra GVN pass in `addMachinePasses` at low opt levels (#43030) | 11 November 2021, 21:40:15 UTC |
6779a9c | Viral B. Shah | 11 November 2021, 20:29:26 UTC | Update suitesparse url (#43046) [ci skip] [skip ci] | 11 November 2021, 20:29:26 UTC |
9c38c20 | woclass | 11 November 2021, 20:15:39 UTC | [doc] Use https for external links whenever possible (#43037) * [doc/dev] Updates link to use `https://` | 11 November 2021, 20:15:39 UTC |
c25b704 | Andreas Varga | 11 November 2021, 10:49:02 UTC | Fix of null column error in geqrf! and gerqf! (#42844) * Fix of null column error in geqrf! and gerqf! * Added tests * Trailing blanks removed * Correct issue number * Adjusted lwork to max(m, 1, Int(real(work[1]))) * Fine tuning | 11 November 2021, 10:49:02 UTC |
017a3af | Sebastian Ament | 11 November 2021, 08:53:02 UTC | Added `factorize` and `logabsdet` for `Number` types (#42912) | 11 November 2021, 08:53:02 UTC |