53401f1 | Keno Fischer | 29 October 2020, 21:36:53 UTC | Improve optimization of harmless call cycles When inference detects a call cycle, one of two things could happen: 1. It determines that in order for inference to converge it needs to limit the signatures of some methods to something more general, or 2. The cycle is determined to be harmless at the inference level (because though there is a cycle in the CFG there is no dependency cycle of type information). In the first case, we simply disable optimizations, which is sensible, because we're likely to have to recompute some information anyway, when we actually get there dynamically. In the second case however, we do do optimizations, but it's a bit of an unusual case. In particular, inference usually delivers methods to inlining in postorder (meaning callees get optimized before their callers) such that a caller can always inline a callee. However, if there is a cycle, there is of course no unique postorder of functions, since by definition we're looking a locally strongly connected component. In this case, we would just essentially pick an arbitrary order (and moreover, depending on the point at which we enter the cycle and subsequently cached, leading to potential performance instabilities, depending on function order). However, the arbitrary order is quite possibly suboptimal. For example in #36414, we have a cycle randn -> _randn -> randn_unlikely -> rand. In this cycle the author of this code expected both `randn` and `_randn` to inline and annotated the functions as such. However, in 1.5+ the order we happed to pick would have inlined randn_unlikely into _randn (had it not been marked noinline), with a hard call break between randn and _randn, whch is problematic from a performance standpoint. This PR aims to address this by introducing a heuristic: If some functions in a cycle are marked as `@noinline`, we want to make sure to infer these last (since they won't ever be inlined anyway). To ensure this happens, while restoring postorder if this happens to break the cycle, we perform a DFS traversal rooted at any `@noinline` functions and then optimize the functions in the cycle in DFS-postorder. Of course still may still not be a true postorder in the inlining graph (if the `@noinline` functions don't break the cycle), but even in that case, it should be no worse than the default order. Fixes #36414 Closes #37234 | 29 October 2020, 21:48:17 UTC |
0d5c38b | Shuhei Kadowaki | 29 October 2020, 16:27:50 UTC | make sure `Random.make_seed()` works on 32bits systems (#38210) | 29 October 2020, 16:27:50 UTC |
0b2c208 | Valentin Churavy | 29 October 2020, 14:51:00 UTC | Merge pull request #38201 from JuliaLang/vc/backtraces insert backtrace printing into user-facing JL_CATCH | 29 October 2020, 14:51:00 UTC |
f613b55 | Daniel Karrasch | 29 October 2020, 13:50:54 UTC | Make ldlt throw with zero pivot (#38051) | 29 October 2020, 13:50:54 UTC |
72971c4 | Michael Abbott | 29 October 2020, 11:34:17 UTC | Let `muladd` accept arrays (#37065) | 29 October 2020, 11:34:17 UTC |
5c47690 | Ian Butterworth | 29 October 2020, 07:56:10 UTC | Add options for varinfo() to show all and imported, sort by each column, search modules recursively (#38042) * add options for varinfo to show all and imported, and sort by size and add option to inspect objects from sub-modules recursively | 29 October 2020, 07:56:10 UTC |
159e6a3 | Jacob Quinn | 29 October 2020, 07:01:57 UTC | Merge pull request #38223 from JuliaLang/jq/printfstr Fix issue where precision > length of string | 29 October 2020, 07:01:57 UTC |
441a9d1 | ScottPJones | 28 October 2020, 18:30:27 UTC | Fix issue where precision > length of string | 29 October 2020, 05:16:59 UTC |
d363fd2 | melonedo | 29 October 2020, 03:41:14 UTC | Fix #36550 (showing SubstitutionString) with the same solution for Regex (#38161) | 29 October 2020, 03:41:14 UTC |
f727f4b | Sacha Verweij | 28 October 2020, 23:50:13 UTC | Merge pull request #38218 from NHDaly/nhd-snoopi_deep-bugfix Fix bug in snoopi_deep: missing copy() on slottypes array | 28 October 2020, 23:50:13 UTC |
4fedd83 | Jeff Bezanson | 28 October 2020, 19:58:34 UTC | add `jit_total_bytes` for getting amount allocated by the JIT (#38030) | 28 October 2020, 19:58:34 UTC |
ea6cdf3 | Nathan Daly | 28 October 2020, 19:49:58 UTC | Fix bug in snoopi_deep: missing copy() on slottypes array The slottypes array is reused and modified, so we need to make a copy of it for each inference frame. This was a bug accidentally introduced in #38143 | 28 October 2020, 19:49:58 UTC |
3c33910 | Valentin Churavy | 27 October 2020, 17:06:59 UTC | Print backtraces from user-facing JL_CATCH Since `jlbacktrace` uses `STDERR_FILENO` directly we have to be a bit careful to avoid mangled errors (particularily on Windows), so we use `STDERR_FILENO` directly, when also using `jlbacktrace`. | 28 October 2020, 19:20:31 UTC |
7f518ef | Jameson Nash | 28 October 2020, 15:59:42 UTC | Fix invmod for BigInt (regression caused by #30515) (#38196) | 28 October 2020, 15:59:42 UTC |
f7dbd1c | Jameson Nash | 28 October 2020, 15:43:18 UTC | inference: mark several no-throw results as pure (#38199) Closes https://github.com/JuliaLang/julia/pull/37230 | 28 October 2020, 15:43:18 UTC |
0f0bc29 | Tim Besard | 28 October 2020, 12:03:01 UTC | Merge pull request #38175 from JuliaLang/tb/reshape_colon Don't call private reshape API in reshape(..., ::Colon). | 28 October 2020, 12:03:01 UTC |
28f68eb | Tim Besard | 28 October 2020, 11:50:57 UTC | Merge pull request #38172 from JuliaLang/tb/jl_get_libllvm Add a function to locate the loaded libLLVM. | 28 October 2020, 11:50:57 UTC |
29a6e66 | Bartosz Białas | 28 October 2020, 07:27:43 UTC | doc: fix broken 'Edit on GitHub' link on release notes page, fixes #34511 (#38191) | 28 October 2020, 07:27:43 UTC |
10b165a | Colin Caine | 28 October 2020, 07:19:05 UTC | manual: fix link in workflow-tips.md (#38205) Bare links are not turned into hyperlinks by Documenter. | 28 October 2020, 07:19:05 UTC |
8f71ea7 | Elliot Saba | 28 October 2020, 02:10:53 UTC | [loading]: Fix new tests post-Preferences merge (#38204) The new Preferences loading strategy changed some pieces of the loading machinery, in particular, it added another parameter to `Base.compilecache_path()`. This was not caught by CI because the Preferences PR stood open for so long that the other tests changed after the CI had already turned green, and we don't re-run tests when another branch is merged, so we often have stale tests. Co-authored-by: Keno Fischer <keno@juliacomputing.com> | 28 October 2020, 02:10:53 UTC |
7e42d51 | Mosè Giordano | 27 October 2020, 23:51:37 UTC | [BinaryPlatforms] More accurate `pentium4` name for base `i686` arch (#38186) | 27 October 2020, 23:51:37 UTC |
4989056 | Logan Kilpatrick | 27 October 2020, 22:53:09 UTC | Update Slack URL (#38187) | 27 October 2020, 22:53:09 UTC |
33b6eda | Stefan Karpinski | 27 October 2020, 21:12:23 UTC | Merge pull request #38190 from JuliaLang/sk/bump-Downloads bump Downloads stdlib (new APIs: progress, request) | 27 October 2020, 21:12:23 UTC |
bd304ef | Elliot Saba | 27 October 2020, 17:46:15 UTC | [loading]: Rework preferences loading (#38044) | 27 October 2020, 17:46:15 UTC |
65c7bf5 | Milan Bouchet-Valat | 27 October 2020, 17:11:44 UTC | Make return type of broadcast inferrable with heterogeneous arrays (#30485) Inference is not able to detect the element type automatically, but we can do it manually since we know promote_typejoin is used for widening. | 27 October 2020, 17:11:44 UTC |
03c5eee | Erik Schnetter | 27 October 2020, 16:56:42 UTC | Prevent integer overflow for large sparse matrices on 32-bit systems. (#37840) | 27 October 2020, 16:56:42 UTC |
861f8d6 | ethomag | 27 October 2020, 16:30:18 UTC | doc manual noteworthy diff from C: division (#38174) | 27 October 2020, 16:30:18 UTC |
1b42549 | Stefan Karpinski | 27 October 2020, 07:12:07 UTC | bump Downloads stdlib (new APIs: progress, request) We also loosen the Base.download tests a bit to allow for more precise exceptions to be raised on error, instead of requiring that ErrorException be thrown. | 27 October 2020, 15:24:07 UTC |
66ac1ea | Stefan Karpinski | 27 October 2020, 15:21:27 UTC | download shim: factor out a little shared logic | 27 October 2020, 15:22:46 UTC |
6312bfe | Jameson Nash | 27 October 2020, 14:55:40 UTC | define and document elsize (#38128) Fixes #36553 Fixes #37438 | 27 October 2020, 14:55:40 UTC |
5d8225a | Jerry Ling | 26 October 2020, 21:18:54 UTC | expose findfirst findnext for UInt8 vector (#37283) * expose findfirst findnext for UInt8 vector * Update base/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * Update base/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * Update base/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * Update base/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * Update base/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * Update test/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * Update base/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * address comments * address comments add OffsetArray test * add findlast findprev * implement comments * let _(r)searchindex handle exception * fix _rsearchindex special behavior * style fix * restirct to 1-indexed array * Update test/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * Update base/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * Update base/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * Update base/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * Update base/strings/search.jl Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> * address comments * change sentinel value to firstindex - 1 * NEWS for find* on Vector of U/Int8 * Update NEWS.md shorten NEWS Co-authored-by: Milan Bouchet-Valat <nalimilan@club.fr> Co-authored-by: Steven G. Johnson <stevenj@mit.edu> | 26 October 2020, 21:18:54 UTC |
c293757 | Simeon Schaub | 26 October 2020, 20:07:24 UTC | fix #37701: support keys(::IOContext) (#37753) | 26 October 2020, 20:07:24 UTC |
c3c4dab | Jeff Bezanson | 26 October 2020, 19:03:58 UTC | fix #37974, struct layout of odd-size primitive types (#38126) | 26 October 2020, 19:03:58 UTC |
f6b51ab | Jeff Bezanson | 26 October 2020, 19:02:41 UTC | fix embedding test (#38117) | 26 October 2020, 19:02:41 UTC |
031c877 | Nathan Daly | 26 October 2020, 18:05:32 UTC | Make `static_show()` print valid identifiers (#38049) | 26 October 2020, 18:05:32 UTC |
da0d968 | Sacha Verweij | 26 October 2020, 17:54:12 UTC | Merge pull request #38171 from JuliaLang/sf/checklist_addendum Note that source distributions should be with USE_BINARYBUILDER=0 | 26 October 2020, 17:54:12 UTC |
6edf6d9 | Simeon Schaub | 26 October 2020, 17:30:53 UTC | allow slurping in lhs of assignment (#37410) fixes #2626 | 26 October 2020, 17:30:53 UTC |
ba39b88 | Bartosz Białas | 26 October 2020, 16:33:10 UTC | fix #37664: treat comments as whitespace (#37729) | 26 October 2020, 16:33:10 UTC |
0bb5626 | Jameson Nash | 26 October 2020, 15:59:44 UTC | fix lineinfo update_line_only compution (#38152) We were emitting extraneous pops when the update should have been trivial (straight code with inlining). | 26 October 2020, 15:59:44 UTC |
2817303 | Tim Besard | 26 October 2020, 11:27:59 UTC | Don't call private reshape API in reshape(..., ::Colon). | 26 October 2020, 11:27:59 UTC |
1246579 | Tim Besard | 26 October 2020, 08:04:14 UTC | Add a function to locate the loaded libLLVM. | 26 October 2020, 09:49:37 UTC |
aa7d201 | Elliot Saba | 26 October 2020, 05:14:53 UTC | Note that source distributions should be with USE_BINARYBUILDER=0 | 26 October 2020, 05:14:53 UTC |
bac321d | Dilum Aluthge | 25 October 2020, 22:53:50 UTC | generate_precompile: don't write to ~/.julia (#38165) | 25 October 2020, 22:53:50 UTC |
3df8f64 | Elliot Saba | 25 October 2020, 19:53:51 UTC | [build] Fix `full-source-dist` downloading LibOSXUnwind on non-OSX platforms (#38047) In `deps/Makefile`, only add `get-osxunwind` to the list of `getall` dependencies if we're running on MacOS. Because we are generally using BinaryBuilder-built dependencies when running `getall`, we'll fail with a 404 error here if we attempt this. | 25 October 2020, 19:53:51 UTC |
a80f903 | Bogumił Kamiński | 25 October 2020, 16:37:36 UTC | Add missing support for single argument *, &, |, xor (#38157) | 25 October 2020, 16:37:36 UTC |
95a0d5d | Rafael Fourquet | 25 October 2020, 09:14:10 UTC | Merge pull request #38098 from JuliaLang/rf/randbang-bigint allow in-place randomization of BigInt and BigFloat | 25 October 2020, 09:14:10 UTC |
d474c98 | Tim Holy | 24 October 2020, 16:32:59 UTC | Inference timing: add a backtrace to head node (#38123) This allows one to identify the caller of runtime-dispatched methods. This can help detect inference problems and know where to fix them. Only "head" nodes (entries into type inference) have backtraces. We avoid doing any processing of the backtrace here, as instruction-pointer lookup is extremely expensive and we don't even know if the information will be used. | 24 October 2020, 16:32:59 UTC |
ae789ca | Keno Fischer | 24 October 2020, 01:55:13 UTC | Fix incorrect fieldtype tfunc (#38148) `fieldtype_tfunc` was returning `Type` as the widest possible return value, but this is not correct, since fields of unwrapped DataTypes can legally be TypeVars. This was in fact causing incorrect execution in the isdefined_tfunc, though on master it happened to accidentally do the correct thing. Fix this by returning `Any` rather than `Type` as the widest type from `fieldtype_tfunc`, but while we're at it, improve the precision of `fieldtype_tfunc` a bit, such that this widening doesn't regress inference precision in some important cases. | 24 October 2020, 01:55:13 UTC |
b8df95f | pixel27 | 23 October 2020, 21:58:34 UTC | println stream should match showerror (#38048) | 23 October 2020, 21:58:34 UTC |
df23697 | Dilum Aluthge | 23 October 2020, 21:23:43 UTC | Bump Pkg from 336af51 to 4a08f33 (#38156) | 23 October 2020, 21:23:43 UTC |
9e993be | Samikshya Chand | 23 October 2020, 21:22:34 UTC | Fix invmod giving wrong results for moduli close to typemax (#30515) Fixes #29971 Co-authored-by: Simon Byrne <simon.byrne@gmail.com> Co-authored-by: sam0410 <samikshya.chand.ece15@iitbhu.ac.in> | 23 October 2020, 21:22:34 UTC |
665279a | Simeon Schaub | 23 October 2020, 21:08:28 UTC | add 'ᵀ postfix operator for transpose (#38062) | 23 October 2020, 21:08:28 UTC |
5faa51b | Jeff Bezanson | 23 October 2020, 21:04:21 UTC | Add AbstractPattern and AbstractMatch to allow for more general pattern matching (#38108) Co-authored-by: ScottPJones <scottjones@alum.mit.edu> | 23 October 2020, 21:04:21 UTC |
1994981 | Jeff Bezanson | 23 October 2020, 21:03:26 UTC | avoid forming circular lower bound in type intersection (#38114) fixes #37685, fixes #38081 | 23 October 2020, 21:03:26 UTC |
bd9abdc | Nathan Daly | 23 October 2020, 19:02:28 UTC | snoopi_deep: Stop stripping function from `slottypes` (#38143) The `slottypes` in a MethodInstance specialization contains the generic function object as the first argument in the array, which we were previously stripping out to "not write redundant information". But this actually makes the output harder to read, because the `show()` for `MethodInstance`s doesn't fully qualify the names of functions. For example, before this commit (note that you might ask "*Which* `fpsort!` is this? What module is it in?"): ```julia 0.001485733 => Core.Compiler.Timings.InferenceFrameInfo(MethodInstance for fpsort!(::Vector{Float64}, ::Base.Sort.QuickSortAlg, ::Base.Order.ForwardOrdering), 0x0000000000007310, Any[], Any[Vector{Float64}, Core.Const(Base.Sort.QuickSortAlg()), Core.Const(Base.Order.ForwardOrdering())]) ``` And after this commit (note that the fully qualified function is the first item in the slottypes array -- it's `Base.Sort.Float.fpsort!`): ```julia 0.001485733 => Core.Compiler.Timings.InferenceFrameInfo(MethodInstance for fpsort!(::Vector{Float64}, ::Base.Sort.QuickSortAlg, ::Base.Order.ForwardOrdering), 0x0000000000007310, Any[], Any[Core.Const(Base.Sort.Float.fpsort!), Vector{Float64}, Core.Const(Base.Sort.QuickSortAlg()), Core.Const(Base.Order.ForwardOrdering())]) ``` While of course you could have gotten this information out of the `MethodInstance` (which is why I left it out), it doesn't show up in the output by default, which makes copy/pasting this output less useful. The output is already quite large, so I don't think the "redundant information" concern is very valid, since it's already quite big, so it's better to include all the necessary information. | 23 October 2020, 19:02:28 UTC |
3e7c8f3 | jmert | 22 October 2020, 18:54:01 UTC | Add predicate-reduce specializations over sparse matrix columns (#37848) | 22 October 2020, 18:54:01 UTC |
31026e8 | Valentin Churavy | 22 October 2020, 17:31:27 UTC | Make Distributed.Worker threadsafe (#38134) Co-authored-by: Jonas Schulze <jonas.schulze@st.ovgu.de> | 22 October 2020, 17:31:27 UTC |
eb64b89 | Jameson Nash | 22 October 2020, 14:37:22 UTC | test: handle runtest errors in a couple more places (#38127) | 22 October 2020, 14:37:22 UTC |
8e3d78d | Jameson Nash | 22 October 2020, 14:36:47 UTC | stacktrace: fix record of inlining frames (#38132) | 22 October 2020, 14:36:47 UTC |
22d50d9 | Alam | 22 October 2020, 12:41:28 UTC | Fixed typo in line 15 (#38103) | 22 October 2020, 12:41:28 UTC |
871ec3f | Michael Abbott | 22 October 2020, 06:36:08 UTC | fix the bracket (#38129) Co-authored-by: Michael Abbott <me@escbook> | 22 October 2020, 06:36:08 UTC |
1c8ab63 | Elliot Saba | 22 October 2020, 01:40:53 UTC | [build] Fix libgit2 build when `USE_BINARYBUILDER=0` (#38121) This fixes a regression in the libgit2 build system which is unable to find the `mbedTLS` headers at compile time. It also synchronizes the `mbedTLS` configuration to include `MD4`, which `libgit2` expects. | 22 October 2020, 01:40:53 UTC |
e3da30c | Valentin Churavy | 21 October 2020, 22:34:37 UTC | Merge pull request #38092 from JuliaLang/vc/llvm12 [LLVM] Update to API changes in LLVM 12 | 21 October 2020, 22:34:37 UTC |
7ac9ac3 | Jameson Nash | 21 October 2020, 21:23:22 UTC | Revert "Make Distributed.Worker threadsafe (#37905)" (#38112) This reverts commit 444aa87348d264fba0cb83486943a97e7db713dd. | 21 October 2020, 21:23:22 UTC |
d288ad8 | Jeff Bezanson | 21 October 2020, 15:43:26 UTC | fix the formatting of some doc strings (#38118) | 21 October 2020, 15:43:26 UTC |
24750c6 | Daniel Karrasch | 21 October 2020, 14:51:34 UTC | Fix tall qr multiplication (#38002) | 21 October 2020, 14:51:34 UTC |
d922cfc | Daniel Karrasch | 21 October 2020, 11:55:55 UTC | Preserve structure under unit[upper/lower]triangular multiplication (#38058) | 21 October 2020, 11:55:55 UTC |
278d4e0 | Fredrik Ekre | 21 October 2020, 09:50:19 UTC | Update doc dependencies, fixes #38116. (#38120) | 21 October 2020, 09:50:19 UTC |
f39cb43 | Michael Jin | 21 October 2020, 02:53:34 UTC | Update source build documentation and Mozilla CA certificate store to latest (10-14-2020) for libgit2 SSL. (#38089) | 21 October 2020, 02:53:34 UTC |
230491b | Elliot Saba | 20 October 2020, 23:00:00 UTC | [build] Add nghttp2 from-source build (#38115) This allows building `nghttp2` when `USE_BINARYBUILDER_NGHTTP2=0`. | 20 October 2020, 23:00:00 UTC |
7a0b1d8 | Fredrik Ekre | 20 October 2020, 22:07:39 UTC | Update Pkg; (#38113) $ git log --pretty=oneline --abbrev=commit fb9c7eb..336af51 336af51e9471ee24d758eaa86cbec48d8a48c991 Precomp: Better error handling, hide cursor during print (#2143) 0dfa1783cdb2da95088c7d5ada67983851a31c71 prevent bad kwargs from being forwarded to precompile (#2142) | 20 October 2020, 22:07:39 UTC |
57f07e3 | Jeff Bezanson | 20 October 2020, 20:51:47 UTC | fix up addition of `¦` and `⌿` operators (#37973) (#38100) | 20 October 2020, 20:51:47 UTC |
fcb2b3d | Thomas van Doornmalen | 20 October 2020, 20:38:10 UTC | jl_llvmpointer_type is a unionall, but jl_is_llvmpointer treats it as a datatype. (#38075) | 20 October 2020, 20:38:10 UTC |
1fcbef9 | Rafael Fourquet | 20 October 2020, 20:35:44 UTC | Random docs: mention StableRNGs in the "Reproducibility" section (#38074) | 20 October 2020, 20:35:44 UTC |
57e6bea | Ravishankar Joshi | 20 October 2020, 17:36:19 UTC | Added tests for checked math functions (#37985) | 20 October 2020, 17:36:19 UTC |
0a1cbec | Kristoffer Carlsson | 20 October 2020, 17:12:04 UTC | fix _crc32c used in loading.jl (#38105) | 20 October 2020, 17:12:04 UTC |
7fa2d39 | Jameson Nash | 20 October 2020, 16:40:09 UTC | Merge pull request #38065 from JuliaLang/jn/mlmatch-faster faster `methods()` | 20 October 2020, 16:40:09 UTC |
6b00c24 | Jameson Nash | 20 October 2020, 16:39:19 UTC | show: skip useless type-aliases (#38099) Fixes #38077 | 20 October 2020, 16:39:19 UTC |
a188457 | Daniel Karrasch | 20 October 2020, 16:37:35 UTC | Fix sparse Subarray fill! (#38021) Co-authored-by: Matt Bauman <mbauman@juliacomputing.com> | 20 October 2020, 16:37:35 UTC |
7cd892f | jaakkor2 | 20 October 2020, 15:09:36 UTC | Prevent scanning of debug files from the current working directory on Windows (#38085) Fixes https://github.com/JuliaLang/julia/issues/36911 [SymSetOptions](https://docs.microsoft.com/en-us/windows/win32/api/dbghelp/nf-dbghelp-symsetoptions) docs say this about SYMOPT_IGNORE_CVREC: > Ignore path information in the CodeView record of the image header when loading a .pdb file. | 20 October 2020, 15:09:36 UTC |
51f8716 | Michael Abbott | 20 October 2020, 09:52:09 UTC | Fine-tune stacktrace printing (#37773) * fine-tune stacktrace printing Co-authored-by: Michael Abbott <me@escbook> | 20 October 2020, 09:52:09 UTC |
7efae4b | Fredrik Ekre | 20 October 2020, 07:17:18 UTC | Update Pkg version. (#38102) $ git log --pretty=oneline --abbrev=commit 0314b79..fb9c7eb fb9c7eb1db30d90f5bfade4dcea2e7c7b1e1c8e7 Reset JULIA_PROJECT environment variable during testing. (#2140) a9a152d08aa4b874613e59376cfbc6bc61cbd4c7 Precompile: Fix and test per-project precomp suspension state retention (#2129) e0539a9d0de08cf3e3da1e8bcb2e9f44435c5381 Delete the `bors.toml` file (#2139) 9d1b2188c11234b57790185b1d17b5b89f94b4d6 Docs: moved Julia 1.4 compat below the section (#1867) 6d2dc3fcefe160cf829fbf5664878733be64816a Print a warning when using the REPL interface in non-interactive mode. (#2076) ac55c0ef35ac80aaef9998376887da1c1c95ee8d Add an example for the SSH-based git protocol. (#2068) 41a3b1204ee9836faf5cce5c255d311974681df6 add a progress bar to precompile process (#2128) 8da19e0f3db3f09239bcb07d9a5dc8339915d61c fixup to progress printing when downloading (#2136) 87873f8037fcd47712578dc0f39eeba9418f9376 send Pkg server value itself as a header to the Pkg server (#2134) 1698ee9377736f10081b63e4ea0997d7dee80640 sort jlls together in status printing (#1987) 88afa647445c65bf50d987f81cfbf85a380618c1 remove using latest registry version when no version in Project (#2133) 5100e0b687236bb24976b3a0dd7e7f8cd4d331bf removed unused argument (#2132) 1e480a5a3185d83abd9dbf3d343452506dd9490b show a progress bar for downloads and build (#2130) | 20 October 2020, 07:17:18 UTC |
8c11d3c | Keno Fischer | 19 October 2020, 22:38:47 UTC | tfuncs: Be more robust in the face of uninhabited types (#37945) Fixes #37943 | 19 October 2020, 22:38:47 UTC |
6329dba | jaakkor2 | 19 October 2020, 22:19:25 UTC | Update python instructions for compiling Julia on Windows (#38088) | 19 October 2020, 22:19:25 UTC |
5f7d06a | Valentin Churavy | 19 October 2020, 02:04:25 UTC | [LLVM] adapt to change in setDefaultFeatures | 19 October 2020, 22:16:00 UTC |
eafb455 | Valentin Churavy | 19 October 2020, 02:02:24 UTC | [LLVM] Implement cl::ParseEnvironmentOptions ourselves | 19 October 2020, 22:16:00 UTC |
96fc147 | Valentin Churavy | 19 October 2020, 01:41:10 UTC | [LLVM] adapt to scalable vector types | 19 October 2020, 22:16:00 UTC |
b08ad36 | Jameson Nash | 19 October 2020, 20:36:23 UTC | Merge pull request #37912 from JuliaLang/jn/37735 Investigate and fix swprintf ccall test failure. Try to ensure the libc runtime is using UTF-8, regardless of the user locale selected. Closes #37735. | 19 October 2020, 20:36:23 UTC |
777f23d | Jakob Nybo Nissen | 19 October 2020, 16:48:13 UTC | Make uniformscaling inequal to matrices (#37525) Closes #35485 | 19 October 2020, 16:48:13 UTC |
c760957 | Pablo Zubieta | 19 October 2020, 16:42:43 UTC | Move command-line options out of Getting Started to separate page, fixes #28845 (#37982) | 19 October 2020, 16:42:43 UTC |
6c8afcb | Shuhei Kadowaki | 19 October 2020, 16:40:33 UTC | doc: add missing triple quotes for `baremodule` doc (#38094) | 19 October 2020, 16:40:33 UTC |
12328e2 | Jameson Nash | 16 October 2020, 20:05:28 UTC | ml-matches: using relaxed sorting when unlimited If we're attempting to apply limits to the results, then it is generally better and faster to use more accurate sorting. Otherwise, approximate sorting should be sufficient and much faster. | 19 October 2020, 16:33:52 UTC |
d65f41d | Jameson Nash | 16 October 2020, 19:58:03 UTC | ml-matches: skip ambiguity work if it won't affect the result | 19 October 2020, 16:10:28 UTC |
ee3ff3c | Jameson Nash | 07 October 2020, 16:35:08 UTC | ensure libc is using a UTF8-compatible encoding Some basic functionality in libc (such as printf) might be broken if the string encoding is not matching our Cstring expectations. | 19 October 2020, 15:34:46 UTC |
3d75d1e | Jameson Nash | 06 October 2020, 21:30:34 UTC | investigate swprintf call failure (#37735) | 19 October 2020, 15:34:46 UTC |
0fcdb73 | Rafael Fourquet | 19 October 2020, 13:54:08 UTC | Random: add a simple SamplerBigInt(r::UnitRange{BigInt}) constructor It has the same effect as calling `Sampler(MersenneTwister, r)`. | 19 October 2020, 13:54:08 UTC |
4cdcc35 | Rafael Fourquet | 19 October 2020, 13:37:12 UTC | allow in-place randomization of BigInt and BigFloat Semantically, `BigInt` and `BigFloat` are immutable, so we don't expose a new API here. But this allows packages knowing what they are doing to not have to duplicate the functionality implemented in `Random`. | 19 October 2020, 13:37:12 UTC |
2655a3a | Keno Fischer | 18 October 2020, 19:49:34 UTC | Drop use of `something` in reflection.jl (#38067) * Drop use of `something` in reflection.jl reflection.jl is included in Core.Compiler, but the `something` function is not, so we cannot use it here. Throw an explicit error instead. I considered having it not throw the error when `err==false`, but it seems to me like passing such a UnionAll is just a user error, rather than an acceptable query, so I left it as is and simply did the bug fix. * Update base/reflection.jl | 18 October 2020, 19:49:34 UTC |
44b55d1 | Kristoffer Carlsson | 18 October 2020, 10:22:31 UTC | make sure TOML.parsefile error when used on a directory (#38078) | 18 October 2020, 10:22:31 UTC |
444aa87 | Jonas Schulze | 17 October 2020, 22:11:25 UTC | Make Distributed.Worker threadsafe (#37905) | 17 October 2020, 22:11:25 UTC |