c0a882a | Simeon David Schaub | 11 November 2021, 18:11:19 UTC | add generic fallback for `eachindex`/`keys` This basically just widens the allowed types for what was previously `EachStringIndex` since the way it works is not really specific to strings. I needed something like this for my own collection type and half expected `eachindex` to work like that already. If people don't like the generic fallback method for `keys` here though, we could also require people to opt into it explicitly by defining `Base.keys(x::T) = Base.EachIndex(x)` themselves instead. | 11 November 2021, 18:20:50 UTC |
13afd74 | woclass | 09 November 2021, 20:55:44 UTC | [doc] Add @id for doc title - Part 2 (base and devdoc) (#43017) * [doc] add `@id base-multimedia` * [doc] add `@id base-keywords` * [doc] add `@id base-big` * [doc] add `@id dev-debug` * [doc] add `@id dev-func-call` * [doc] add `@id dev-stdio-libuv` * [doc] add `@id dev-sysimg-build` * [doc] add `@id dev-locks` * [doc] add `@id dev-stdio-ios-c` | 09 November 2021, 20:55:44 UTC |
7e85914 | Jeff Bezanson | 09 November 2021, 18:44:37 UTC | document compatibility properties of `serialize` (#41520) | 09 November 2021, 18:44:37 UTC |
1cda795 | Jeff Bezanson | 09 November 2021, 18:44:00 UTC | export `jl_timing_data` for separate codegen lib (#43002) Fixes ENABLE_TIMINGS option | 09 November 2021, 18:44:00 UTC |
e3a8ba2 | Shuhei Kadowaki | 09 November 2021, 18:43:35 UTC | inference: normalize some internal naming (#42987) Particularly normalize some names with the following schema: - `atype`: signature type - `argtypes`: inference lattice elements of call arguments Also include very minor cleanups. | 09 November 2021, 18:43:35 UTC |
ab0c6dd | Pepijn de Vos | 09 November 2021, 18:41:54 UTC | Add missing type parameter to TakeWhile (#42958) Without this `TakeWhile` has `eltype` of `Any` | 09 November 2021, 18:41:54 UTC |
0771250 | Jameson Nash | 09 November 2021, 18:40:53 UTC | Merge pull request #42962 from JuliaLang/jn/opts-internals | 09 November 2021, 18:40:53 UTC |
f458b07 | Harmen Stoppels | 09 November 2021, 13:52:25 UTC | Add missing USE_SYSTEM_LIBWHICH initialization (#43014) | 09 November 2021, 13:52:25 UTC |
d87c80c | Rafael Fourquet | 09 November 2021, 13:23:01 UTC | fix show for RandomDevice (#41233) * fix show for RandomDevice On Windows, this was displaying an internal buffer, and on Linux, this was showing the keyword argument `unlimited` as if it was a positional argument. * rm spurious space Co-authored-by: Simeon Schaub <simeondavidschaub99@gmail.com> * Update stdlib/Random/test/runtests.jl Co-authored-by: Simeon Schaub <simeondavidschaub99@gmail.com> * Update stdlib/Random/test/runtests.jl Co-authored-by: Simeon Schaub <simeondavidschaub99@gmail.com> Co-authored-by: Simeon Schaub <simeondavidschaub99@gmail.com> | 09 November 2021, 13:23:01 UTC |
8fc7e1b | Dilum Aluthge | 09 November 2021, 05:40:45 UTC | FileWatching: fix typo on #42926 (2539a679ab) (#43010) Co-authored-by: Jameson Nash <vtjnash@gmail.com> | 09 November 2021, 05:40:45 UTC |
067ac5e | Jameson Nash | 08 November 2021, 22:17:41 UTC | FileWatching: fix typo on #42926 (2539a679ab) | 08 November 2021, 22:17:41 UTC |
cfb7343 | Jameson Nash | 05 November 2021, 16:49:27 UTC | add more info to SIGUSR1 printing This is similar to what the kernel prints for us when sent SIGINFO, and can be sometimes useful for getting the pid of a process, and other status details. | 08 November 2021, 21:57:55 UTC |
e325497 | Jameson Nash | 03 November 2021, 20:15:20 UTC | use more efficient calling convention for calls with few arguments | 08 November 2021, 21:57:55 UTC |
340982f | Jameson Nash | 03 November 2021, 19:18:08 UTC | codegen: fix needloop for replacefield Wrong previously for String and SimpleVector, and also less optimal than was feasible | 08 November 2021, 21:57:55 UTC |
fc565a4 | Jameson Nash | 03 November 2021, 19:17:20 UTC | improve jl_pointer_egal for more cases | 08 November 2021, 21:57:55 UTC |
83b7dbb | Jameson Nash | 08 November 2021, 21:43:26 UTC | atomics: fix replacefield loop We would previously loop back to the wrong point in the code, and cause potentially chaos and malformed IR. | 08 November 2021, 21:45:50 UTC |
d6f59fa | Jameson Nash | 08 November 2021, 19:49:17 UTC | Timer: handle timeout correctly (#42854) I am not sure why we ever used round+1 instead of ceil+1, as this is simply strictly more correct. | 08 November 2021, 19:49:17 UTC |
27f579f | Jameson Nash | 08 November 2021, 19:45:42 UTC | inference: relax type_more_complex for Type (#42901) If the object is not a potential Type{...}, we can allow greater flexibility in recursion. | 08 November 2021, 19:45:42 UTC |
2539a67 | Jameson Nash | 08 November 2021, 19:43:03 UTC | FileWatching: fix API problems with FDWatcher (#42926) * add remaining FDEvents, and modernize this struct * add `isopen` function * do not deliver events after the FDWatcher is closed * do not delete events that have not been delivered * support running of these tests with Revise.jl * note: some threading problems still exist (where marked) | 08 November 2021, 19:43:03 UTC |
b6eb87e | Valentin Churavy | 08 November 2021, 19:38:53 UTC | Merge pull request #42963 from JuliaLang/jn/header-cleanup Progress towards header cleanup | 08 November 2021, 19:38:53 UTC |
ee4bcab | Harmen Stoppels | 08 November 2021, 19:30:50 UTC | Allow system versions of libwhich and libblastrampoline (#43000) | 08 November 2021, 19:30:50 UTC |
ca28619 | Takafumi Arakaki | 08 November 2021, 19:29:53 UTC | Add JULIA_THREAD_SLEEP_THRESHOLD environment variable (#42981) This patch re-introduces the code originally added in 4752908e61f45f85b684b2c8ed0085f0cf37f416 and was removed in 93e3d284a6205d9a9cca6d5bdb3856439e0b55ad. | 08 November 2021, 19:29:53 UTC |
12b9bec | Takafumi Arakaki | 08 November 2021, 17:06:02 UTC | Fix segfault while profiling task switching (#42973) | 08 November 2021, 17:06:02 UTC |
d2b5a13 | Stefan Karpinski | 08 November 2021, 16:16:45 UTC | DiskStat: show only virtual fields (#42959) Also, return them as Int64 since that's what we use in Julia land for sizes of things. The raw fields remain UInt64 because that's how they are actually represented. | 08 November 2021, 16:16:45 UTC |
d39b2c0 | Dilum Aluthge | 08 November 2021, 13:13:52 UTC | CI (Buildkite): increase the timeout of the code coverage job (#42994) | 08 November 2021, 13:13:52 UTC |
e3b9290 | Ian Butterworth | 08 November 2021, 03:15:28 UTC | Revert "Add partial isequal for VersionNumber (#42780)" (#42992) | 08 November 2021, 03:15:28 UTC |
fa6dcb0 | Alex Arslan | 08 November 2021, 01:46:27 UTC | Keep expected crash in misc test from producing a core dump (#42990) Currently the test in `test/misc.jl` that tests that a particular kind of read fault crashes has two issues: 1. It works for the wrong reason. It's testing for `!success` when evaluating code interpolated directly into the `Cmd`, but it isn't quoting the code, so the subprocess fails with a syntax error instead of the type of crash the test is expecting. 2. On some platforms (such as our good ol' pal FreeBSD), this kind of read fault produces a core dump. This a bit annoying because it means that the repo state becomes dirty after running the tests and it may overwrite an existing core dump that was left behind by an earlier issue we'd like to diagnose. To fix these, we can wrap the code passed to the subprocess in single quotes and on Unix-like systems wrap the subprocess in `ulimit -c 0` to avoid producing a core dump. | 08 November 2021, 01:46:27 UTC |
aca020b | DilumAluthgeBot | 07 November 2021, 20:33:29 UTC | 🤖 Bump the Downloads stdlib from 4250b35 to d8c626b (#42980) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 07 November 2021, 20:33:29 UTC |
56f54e4 | DilumAluthgeBot | 07 November 2021, 20:33:14 UTC | 🤖 Bump the Pkg stdlib from 66d98f7b to 17df5a4f (#42984) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 07 November 2021, 20:33:14 UTC |
c1f94b6 | Ian Butterworth | 07 November 2021, 20:17:07 UTC | Add partial isequal for VersionNumber (#42780) | 07 November 2021, 20:17:07 UTC |
310de1c | Shuhei Kadowaki | 07 November 2021, 13:39:33 UTC | optimizer: enhance SROA, handle partially-initialized allocations (#42834) During adding more test cases for our SROA pass, I found our SROA doesn't handle allocation sites with uninitialized fields at all. This commit is based on #42833 and tries to handle such "unsafe" allocations, if there are safe `setfield!` definitions. For example, this commit allows the allocation `r = Ref{Int}()` to be eliminated in the following example (adapted from <https://hackmd.io/bZz8k6SHQQuNUW-Vs7rqfw?view>): ```julia julia> code_typed() do r = Ref{Int}() r[] = 42 b = sin(r[]) return b end |> only ``` This commit comes with a plenty of basic test cases for our SROA pass also. | 07 November 2021, 13:39:33 UTC |
71d57d9 | Shuhei Kadowaki | 07 November 2021, 13:38:52 UTC | inference: eliminate more potential `widenconst(::TypeofVararg)` cases (#42971) Follows #42583. The missing case within `const_prop_function_heuristic` was originally reported from RerationalAI's test suite. I added the other two cases by some code review. | 07 November 2021, 13:38:52 UTC |
36304ba | Dilum Aluthge | 07 November 2021, 07:04:30 UTC | Running `make testall` should run all of the tests (#42904) | 07 November 2021, 07:04:30 UTC |
4304c51 | Dilum Aluthge | 07 November 2021, 06:52:08 UTC | CI (Buildkite): fix another bug in the once-daily scheduled Buildkite job (#42983) | 07 November 2021, 06:52:08 UTC |
9bb8a6f | Dilum Aluthge | 07 November 2021, 06:51:39 UTC | CI (GHA, Buildbot): Update the list of commit statuses created by the "Create Buildbot Statuses" workflow (#42985) | 07 November 2021, 06:51:39 UTC |
f728038 | Dilum Aluthge | 07 November 2021, 06:51:01 UTC | CI (Buildkite): increase the timeouts for the Buildkite jobs that run under `rr` (#42979) | 07 November 2021, 06:51:01 UTC |
f3d505d | Dilum Aluthge | 07 November 2021, 00:59:43 UTC | CI (Buildkite): Fix the Buildkite once-daily scheduled job (#42976) | 07 November 2021, 00:59:43 UTC |
96a99c2 | Alex Arslan | 07 November 2021, 00:56:34 UTC | Fix libunwind segfaults from JIT frames on FreeBSD (#42970) This adds libunwind PR 308 as a patch and updates LibUnwind_jll to use a version with that patch included. | 07 November 2021, 00:56:34 UTC |
653cad3 | Alex Arslan | 06 November 2021, 20:25:32 UTC | Partially revert #42919 (#42969) * Partially revert #42919 The change to RPATH setting for macOS seems to have improved the situation for users on Monterey, but the other changes in #42919 that affect other platforms seem to have caused regressions on musl Linux (#42940) and FreeBSD (#42944). The easiest thing to do here is to keep the changes from #42919 as they apply to macOS but revert the others. * Un-revert the `RPATH_LIB` definition | 06 November 2021, 20:25:32 UTC |
47255f9 | Sebastian Ament | 06 November 2021, 19:56:42 UTC | added logabsdet for Cholesky factorizations (#42910) | 06 November 2021, 19:56:42 UTC |
a492d96 | N5N3 | 06 November 2021, 19:32:25 UTC | Make UniformScaling's getindex respect inputs' axes (#42886) | 06 November 2021, 19:32:25 UTC |
2ece1cd | Guillaume Berger | 06 November 2021, 18:04:39 UTC | Fix error in the doc of `nullspace` (#42960) | 06 November 2021, 18:04:39 UTC |
dccf0af | Kenta Sato | 06 November 2021, 17:53:09 UTC | doc: add missing . in sqrt(::AbstractMatrix) docstring (#42956) | 06 November 2021, 17:53:09 UTC |
4c6696e | Shuhei Kadowaki | 06 November 2021, 08:08:37 UTC | optimizer: more improve the idempotency of callsite inlining (#42933) We need to force constant propagation of call that is going to be force-inlined by callsite inlining annotation, since it needs to find that constant result if the constant arguments arrive there. The added test case should describe the motivation. | 06 November 2021, 08:08:37 UTC |
649573c | Oscar Smith | 05 November 2021, 20:38:00 UTC | revert fma tests on windows (#42955) | 05 November 2021, 20:38:00 UTC |
c5a7d02 | Shuhei Kadowaki | 05 November 2021, 20:11:17 UTC | optimizer: more refactors on inlining tests (#42932) | 05 November 2021, 20:11:17 UTC |
c12aca8 | Jameson Nash | 05 November 2021, 18:09:06 UTC | remove uv.h from julia.h It is still installed, if needed, though we had already stopped exporting its functions anyways. | 05 November 2021, 18:51:25 UTC |
439816a | Jameson Nash | 05 November 2021, 17:16:52 UTC | remove some unnecessary header content and exports | 05 November 2021, 18:43:06 UTC |
7368c49 | Jameson Nash | 05 November 2021, 18:30:59 UTC | fix missing function declaration for gcext | 05 November 2021, 18:43:06 UTC |
7d41d1e | Dilum Aluthge | 05 November 2021, 12:23:30 UTC | CI (Buildkite): increase the default timeouts (#42953) | 05 November 2021, 12:23:30 UTC |
b777d02 | Dilum Aluthge | 05 November 2021, 11:27:29 UTC | CI (GHA, Buildbot): temporarily, don't create the `linuxaarch64` Buildbot status (#42954) | 05 November 2021, 11:27:29 UTC |
7c9d126 | Dilum Aluthge | 05 November 2021, 08:52:51 UTC | Distributed test suite: mark another test as thread-unsafe (#42941) | 05 November 2021, 08:52:51 UTC |
fa62616 | DilumAluthgeBot | 05 November 2021, 08:52:23 UTC | 🤖 Bump the Downloads stdlib from c91876a to 4250b35 (#42935) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 05 November 2021, 08:52:23 UTC |
d00d457 | Tim Holy | 04 November 2021, 23:08:10 UTC | Docs: `TwicePrecision` is an internal type (#42863) Fixes #42857 | 04 November 2021, 23:08:10 UTC |
486c8de | Elliot Saba | 04 November 2021, 21:08:29 UTC | Eliminate `@loader_path/julia` element of `DL_LOAD_PATH` (#42930) Analogous to https://github.com/JuliaLang/julia/commit/236523ffc2b0493e8e519a721ec030b7c3a64f83, we don't need to search in `@loader_path/julia` anymore, as `libjulia-internal` is now consistently within the same directory as our other dependencies. | 04 November 2021, 21:08:29 UTC |
8e87263 | Dilum Aluthge | 04 November 2021, 19:28:51 UTC | CI (Buildkite): add Pkg, Artifacts, and LazyArtifacts to the list of "network-related stdlibs" that we test in a separate Buildkite job (#42885) | 04 November 2021, 19:28:51 UTC |
707c57c | Oscar Smith | 04 November 2021, 14:22:22 UTC | fixed Float16 from Float64 and BigFloat (#42837) * fixed Float16 from Float64 and BigFloat. Many thanks to Jamison. | 04 November 2021, 14:22:22 UTC |
ee36c13 | Oscar Smith | 04 November 2021, 14:13:48 UTC | Emulated `fma` (#42783) Emulated `fma` for cases when hardware fma is not available. Generally pre-Haswell, some arm, etc. Co-authored-by: oscarddssmith <oscar.smith@juliacomputing.com> | 04 November 2021, 14:13:48 UTC |
e7df4a6 | DilumAluthgeBot | 04 November 2021, 13:38:14 UTC | 🤖 Bump the Pkg stdlib from 570b6058 to 66d98f7b (#42927) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 04 November 2021, 13:38:14 UTC |
236523f | Elliot Saba | 03 November 2021, 17:16:44 UTC | Cleanup `RPATH` settings (#42919) Previously, we needed to provide both `usr/lib` and `usr/lib/julia` as elements on the `RPATH`, because `libjulia` lived in `usr/lib`, and needed to be able to find dependent libraries both within a build tree (when most libraries live in `usr/lib`) and within an install tree (when most libraries live in `usr/lib/julia`). Nowadays, everything is either in `usr/lib` or in `usr/lib/julia` with the exception of `libjulia`, but since it's really `libjulia-internal` that does all the `dlopen()`'ing, we can simply use `$$ORIGIN` (or the equivalent `@loader_path/` on macOS) and completely ignore the rest of the RPATH shenanigans we do. | 03 November 2021, 17:16:44 UTC |
b1cf46c | DilumAluthgeBot | 03 November 2021, 04:56:59 UTC | 🤖 Bump the Tar stdlib from e65daff to 6a94602 (#42914) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 03 November 2021, 04:56:59 UTC |
e8cebcd | Johnny Chen | 03 November 2021, 04:13:37 UTC | docs: add some missing backquote to cmd options (#41869) * docs: add some missing backquote to cmd options <skip ci> * also change the phrase | 03 November 2021, 04:13:37 UTC |
f64c1ee | Sacha Verweij | 03 November 2021, 04:04:30 UTC | Add patch release contribution docs to CONTRIBUTING.md. (#39357) * Add patch release contribution docs to CONTRIBUTING.md. Co-authored-by: Alex Arslan <ararslan@comcast.net> Co-authored-by: Viral B. Shah <ViralBShah@users.noreply.github.com> | 03 November 2021, 04:04:30 UTC |
40772d0 | Likan Zhan | 03 November 2021, 03:54:23 UTC | Update integers-and-floating-point-numbers.md (#42755) * Update integers-and-floating-point-numbers.md This sentence also confuses me. Copied from [StefanKarpinski](https://discourse.julialang.org/t/i-can-not-understand-the-meaning-of-this-sentence-on-the-doc/21010) * Update integers-and-floating-point-numbers.md * fix whitespace Co-authored-by: Viral B. Shah <ViralBShah@users.noreply.github.com> | 03 November 2021, 03:54:23 UTC |
79b0226 | Logan Kilpatrick | 02 November 2021, 22:29:48 UTC | Update index.md (#42234) | 02 November 2021, 22:29:48 UTC |
2109c86 | Katharine Hyatt | 02 November 2021, 22:26:29 UTC | Two small doc nits (#42769) * Doctest for elsize * xref to notify Co-authored-by: Simeon David Schaub <schaub@mit.edu> | 02 November 2021, 22:26:29 UTC |
d9b1a3c | Jeff Bezanson | 02 November 2021, 19:52:44 UTC | add `--strip-metadata` option (#42513) | 02 November 2021, 19:52:44 UTC |
e03ead0 | Dilum Aluthge | 02 November 2021, 18:29:03 UTC | Remove some other uses of the unauthenticated `git://` protocol in Git URLs (#42907) | 02 November 2021, 18:29:03 UTC |
7eba9c1 | Dilum Aluthge | 02 November 2021, 17:54:07 UTC | External stdlibs: stop using the unauthenticated `git://` protocol in the external stdlib URLs (#42906) | 02 November 2021, 17:54:07 UTC |
c646b5d | Daniel Karrasch | 02 November 2021, 13:56:35 UTC | Specialize findmax/findmin on SparseVector, fixes #42823 (#42825) | 02 November 2021, 13:56:35 UTC |
2179795 | Daniel Karrasch | 02 November 2021, 10:17:54 UTC | Simplify minimum/maximum on sparse vectors (#42845) | 02 November 2021, 10:17:54 UTC |
a0ff944 | Daniel Karrasch | 02 November 2021, 07:42:35 UTC | Cleanup imports in SparseArrays (#42894) | 02 November 2021, 07:42:35 UTC |
8544c5a | Daniel Karrasch | 02 November 2021, 07:35:50 UTC | Add `sortby` keyword to `eigvals` for SymOrHerm (#42900) * Add `sortby` keyword to `eigvals` for SymOrHerm * fix sort | 02 November 2021, 07:35:50 UTC |
b4552e7 | Mark Kittisopikul | 02 November 2021, 05:35:44 UTC | Add hint for :quit to REPL.REPLCompletions.UndefVarError_hint (#41990) | 02 November 2021, 05:35:44 UTC |
12b2ec7 | Dilum Aluthge | 02 November 2021, 05:32:55 UTC | `choosetests`: `Pkg/pkg` is no longer relevant (#42890) | 02 November 2021, 05:32:55 UTC |
b72d552 | DilumAluthgeBot | 02 November 2021, 05:18:10 UTC | 🤖 Bump the Pkg stdlib from 972fe337 to 570b6058 (#42897) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 02 November 2021, 05:18:10 UTC |
5d80a87 | DilumAluthgeBot | 02 November 2021, 05:17:02 UTC | 🤖 Bump the SHA stdlib from c5dd533 to d30dbf6 (#42895) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 02 November 2021, 05:17:02 UTC |
5ffee12 | Ian Butterworth | 02 November 2021, 03:57:14 UTC | Profile: metadata handling tweaks (#42868) As was discussed in #42482 after merge, there's some things that could be done better: - Reverts the is_block_end logic, given it should only be run on data that includes metadata, so can afford to be more efficient - Adds has_meta for identifying when profile data has metadata - Adds metadata checks to functions that expect metadata to be present - Changes Profile.fetch to by default include metadata. The reason for this is that it was previously made to strip metadata to avoid breaking existing downstream Profile data consumers, but it turns out that they use Profile internals like tree! that require the metadata to be present, so this ended up causing an unintended breakage anyway - Adds consts for the metadata field offsets, for consumers to use | 02 November 2021, 03:57:14 UTC |
52ff195 | Elliot Saba | 01 November 2021, 20:47:23 UTC | Fix no-codegen test on musl64 tester (#42827) Apparently, busybox has strange behavior when you copy `$pfx/..` and the destination directory exists: ``` $ docker run -ti alpine # mkdir -p /tmp/foo/bar; touch /tmp/foo/bar/a /tmp/foo/bar/b # cp -vr /tmp/foo/bar/.. /tmp/foo2 '/tmp/foo/bar/../bar/a' -> '/tmp/foo2/bar/a' '/tmp/foo/bar/../bar/b' -> '/tmp/foo2/bar/b' '/tmp/foo/bar/../bar' -> '/tmp/foo2/bar' '/tmp/foo/bar/..' -> '/tmp/foo2' # cp -vr /tmp/foo/bar/.. /tmp/foo2 '/tmp/foo/bar/../bar/a' -> '/tmp/foo2/../bar/a' '/tmp/foo/bar/../bar/b' -> '/tmp/foo2/../bar/b' '/tmp/foo/bar/../bar' -> '/tmp/foo2/../bar' '/tmp/foo/bar/..' -> '/tmp/foo2/..' ``` We'll dodge this by using Julia's `cp()` function. * Eliminate usage of `rm` command-line program Note that we use `recursive=true` as we might encounter a `libjulia-codegen.X.Y.dylib.dSYM` directory. Although deleting that directory is not critical for this test, not throwing an exception is, so we just remove it as well. Also, this test would silently fail to test what we want it to if run out of a build directory, so we first teach it to dynamically find the location of `libjulia-codegen`, then secondly, cause it to fail if it does not remove any `libjulia-codegen` libraries. | 01 November 2021, 20:47:23 UTC |
f9bb6f8 | Christopher Burns | 01 November 2021, 19:08:43 UTC | make the spacing between suggested completions consistent (#42891) | 01 November 2021, 19:08:43 UTC |
31b9fd2 | Dilum Aluthge | 01 November 2021, 17:46:21 UTC | `choosetests`: add the `--force-net` option, which will throw an error if networking is unavailable (#42889) | 01 November 2021, 17:46:21 UTC |
cae3571 | DilumAluthgeBot | 01 November 2021, 12:24:51 UTC | 🤖 Bump the Pkg stdlib from 01e9121d to 972fe337 (#42888) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 01 November 2021, 12:24:51 UTC |
a121721 | Shuhei Kadowaki | 01 November 2021, 10:49:07 UTC | inference: form `PartialStruct` for extra type information propagation (#42831) * inference: form `PartialStruct` for extra type information propagation This commit forms `PartialStruct` whenever there is any type-level refinement available about a field, even if it's not "constant" information. In Julia "definitions" are allowed to be abstract whereas "usages" (i.e. callsites) are often concrete. The basic idea is to allow inference to make more use of such precise callsite type information by encoding it as `PartialStruct`. This may increase optimization possibilities of "unidiomatic" Julia code, which may contain poorly-typed definitions, like this very contrived example: ```julia struct Problem n; s; c; t end function main(args...) prob = Problem(args...) s = 0 for i in 1:prob.n m = mod(i, 3) s += m == 0 ? sin(prob.s) : m == 1 ? cos(prob.c) : tan(prob.t) end return prob, s end main(10000, 1, 2, 3) ``` One of the obvious limitation is that this extra type information can be propagated inter-procedurally only as a const-propagation. I'm not sure this kind of "just a type-level" refinement can often make constant-prop' successful (i.e. shape-up a method body and allow it to be inlined, encoding the extra type information into the generated code), thus I didn't not modify any part of const-prop' heuristics. So the improvements from this change might not be very useful for general inter-procedural analysis currently, but they should definitely improve the accuracy of local analysis and very simple inter-procedural analysis. | 01 November 2021, 10:49:07 UTC |
6c274ed | Shuhei Kadowaki | 01 November 2021, 06:24:10 UTC | optimizer: remove unnecessary checks (#42884) Now we never form `ConstCallInfo(::InvokeCallInfo, results)` and so the `sig.f === Core.invoke` check is no longer needed. This commit also abstracts the common `Core.invoke` rewrite pattern into `invoke_rewrite(::Vector{Any})` utility. | 01 November 2021, 06:24:10 UTC |
3863631 | Valentin Churavy | 01 November 2021, 05:07:18 UTC | Update patchelf.mk to always use bzp2 (#42881) | 01 November 2021, 05:07:18 UTC |
d7c28c8 | DilumAluthgeBot | 01 November 2021, 04:15:30 UTC | 🤖 Bump the Pkg stdlib from 26918395 to 01e9121d (#42883) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 01 November 2021, 04:15:30 UTC |
8d82176 | Paul | 01 November 2021, 03:12:52 UTC | use the httable key instead of the binding name in jl_module_names (#42532) | 01 November 2021, 03:12:52 UTC |
ad607a2 | Jameson Nash | 01 November 2021, 03:05:31 UTC | handle some cases of null Task in segv handling (#42836) Also ensure we do not use round-robin sampling when printing a critical error, only for profiling. | 01 November 2021, 03:05:31 UTC |
b55fb5f | Dilum Aluthge | 01 November 2021, 02:06:21 UTC | `choosetests`: preparations for setting up a separate Buildkite job that runs the `Pkg` test suite (#42859) | 01 November 2021, 02:06:21 UTC |
9ac5582 | Dilum Aluthge | 31 October 2021, 23:43:28 UTC | Test suite: improve the warning that we print when we skip the `Profile` tests on ARM (#42862) | 31 October 2021, 23:43:28 UTC |
1fc5c8c | DilumAluthgeBot | 31 October 2021, 23:37:51 UTC | 🤖 Bump the LibCURL stdlib from cddeb7f to 04c450c (#42869) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 31 October 2021, 23:37:51 UTC |
636a136 | DilumAluthgeBot | 31 October 2021, 23:35:29 UTC | 🤖 Bump the SuiteSparse stdlib from b15c39b to e4df734 (#42872) Co-authored-by: Dilum Aluthge <dilum@aluthge.com> | 31 October 2021, 23:35:29 UTC |
6f55bf9 | Kristoffer Carlsson | 31 October 2021, 18:08:00 UTC | no longer explicitly print some things in white (#42864) | 31 October 2021, 18:08:00 UTC |
ca6b3ba | Sacha Verweij | 31 October 2021, 18:04:11 UTC | Optimize show(io::IO, m::Module) implementation. (#42773) show(io::IO, m::Module) allocates. This commit provides an implementation that does not allocate, improving perf. | 31 October 2021, 18:04:11 UTC |
86f656d | Daniel Karrasch | 31 October 2021, 12:23:06 UTC | Allow vectors in reflectorApply! (#42874) Co-authored-by: Sheehan Olver <solver@mac.com> | 31 October 2021, 12:23:06 UTC |
3003742 | Dilum Aluthge | 31 October 2021, 08:23:15 UTC | CI (Buildkite): add a separate Buildkite job that runs the tests of the network-related stdlibs (e.g. Downloads.jl), and automatically retries that job up to a maximum number of tries (#42861) | 31 October 2021, 08:23:15 UTC |
3ed326a | Valentin Churavy | 31 October 2021, 00:18:47 UTC | Merge pull request #42800 from JuliaLang/vc/patchelf Update Patchelf | 31 October 2021, 00:18:47 UTC |
66d05d5 | N5N3 | 30 October 2021, 16:28:38 UTC | Add missing optimization for `*`,`/` between Diagonal and Triangular (#42343) | 30 October 2021, 16:28:38 UTC |
4a12d1e | Malte Sandstede | 30 October 2021, 11:48:47 UTC | Add diskstat() function returning statistics of the disk. (#42248) | 30 October 2021, 11:48:47 UTC |
a387724 | Shuhei Kadowaki | 30 October 2021, 08:41:05 UTC | inference: wrap constant-prop' result directly within `OpaqueClosureCallInfo` (#42849) This is more consistent with the arrangement of `InvokeCallInfo`, and will make succeeding processing a bit cleaner. This change also avoids unnecessary specializations in callinfo constructions. | 30 October 2021, 08:41:05 UTC |