(cherry picked from commit e96b19d)

* Allow `BUILDKITE_BRANCH` to provide branch name

Our CI system checks commits out as a detached head, which breaks our
`Base.GIT_VERSION_INFO.branch` information.

* Fix typo

(cherry picked from commit d117975)

…42574)

Co-authored-by: Daniel Karrasch <[email protected]>
(cherry picked from commit b3c268c)

`b->name` is used to lookup in `b->owner`, not `var`.

(cherry picked from commit dbe41d4)

…by up to one microsecond (to compensate for Windows tar) (#45552)

(cherry picked from commit 4c39647)

…s do not overlap. (#45578)

* Fix union!(s::BitSet, r::AbstractUnitRange{<:Integer}) when two ranges do not overlap.
Resizing of BitSet should be filled with 0 by default.

(cherry picked from commit 5e8e0a5)

Co-authored-by: Viral B. Shah <[email protected]>
(cherry picked from commit 5811825)

(cherry picked from commit 25e3809)

(cherry picked from commit 7beeaf7)

When trying to dlopen a file with non-standard extension (e.g. `foo.so`
instead of `foo.dylib` when running on macOS), if this failed (e.g.
because of a reference to an undefined symbol), then instead of printing
the error message returned by `dlerror` with a helpful notice what
went wrong, a message indicating something to the effect that
"foo.so.dylib was not found" was shown, which was not helpful at all.

To get the actual helpful error message, add a check so that when dlopen
fails for a file that actually exists, we don't retry loading from a
file with the standard extension attached, which might not even exist;
instead we just give up.

This matches what is already being done for relative paths. This patch
simply copies the relevant check to also be applied to the case dealing
with absolute paths.

(cherry picked from commit a490197)

…hefile (#47433)

* CircleCI caching: allow ftime to be ceil(ftime_req) in Base.stale_cachefile

It appears that [caching functionalities](https://circleci.com/docs/caching/) provided by CircleCi, a leading CI/CD provider, can truncate timestamps to full seconds, resulting in re-compilations as below:
```
Rejecting stale cache file /root/.julia/compiled/v1.8/ComponentArrays/cYHSD_3rQji.ji (mtime 1.6673960929277816e9) because file /root/.julia/packages/ComponentArrays/YyD7i/src/ComponentArrays.jl
```

This PR relaxes the `is_stale` check to be robust against rounding-to-second timestamp mutations.

I can provide a minimal CircleCI configuration file to reproduce if this is helpful.

(cherry picked from commit bf92e83)

[PCRE2_jll] Upgrade to 10.42

(cherry picked from commit 1c7cdd5)

(cherry picked from commit ef511c9)

as reported in https://discourse.julialang.org/t/indexed-assignment-with-logical-indices-subarray-jl-error/83646

(cherry picked from commit 89bdcce)

(cherry picked from commit b64743b)

This fixes some issues around macro hygiene in `global` expressions.
Apparently we always treat l-values in global expressions as being
escaped, but we still need to be careful to handle type annotations and
destructuring correctly.

(cherry picked from commit cc25a13)

This code was introduced by me back in #31025 to speed up evaluation
of generated functions that didn't make use of all of their arguments to
make generation decisions. However, it neglected to take into account the
possibility that the generator could be varargs. As a result, an unfortunate
coincidence of an unused slot in the correct position could have allowed
expansion of generators that were not supposed to be expandable. This can
cause incorrect inference with all the usual consequences. However, fortunately
this coincidence appears to be pretty rare.

Fixes JuliaDebug/CassetteOverlay.jl#12

(cherry picked from commit 328dd57)

LLVM adds the nonnull attribute on its own, which makes
the verifier fail.

Fixes #47245

(cherry picked from commit bc39fd1)

Different platforms implement qsort differently, leading to
platform-specific errors. This is a quick port of the ml_matches
algorithm for use instead. For small unions (almost always), this should
also be slightly faster, though insignificant.

Refs #45874

(cherry picked from commit 8cc5445)

(cherry picked from commit fd66c30)

(cherry picked from commit 8003563)

…40136)

(cherry picked from commit fbd718f)

(cherry picked from commit 3661a08)

(cherry picked from commit 7514bcf)

(cherry picked from commit e6d9979)

(cherry picked from commit 73c1eeb)

* bugfix: fld1 order of ops

fixes #28973

(cherry picked from commit fcdc5bc)

1. map `copyto!(::BitArray, n1, ::BitArray, n2, l)` to `Base.unsafe_copyto!`
2. add missing unaliasing in `copyto!` for `AbstractArray`

(cherry picked from commit 0ea2b2d)

* Fix GC assertion on array of derived pointers

* Add test

(cherry picked from commit d885fc7)

(cherry picked from commit 3c04919)

* Fix the issue #43921

* add a test

Co-authored-by: Kristoffer <[email protected]>
(cherry picked from commit b8a77da)

(cherry picked from commit da13d78)

This should prevent LinearAlgebra from trying to increase our OpenBLAS
thread count in its `__init__()` method when we're not trying to enable
threaded BLAS.

(cherry picked from commit a8b3994)
(cherry picked from commit 8830c26)

Co-authored-by: Jameson Nash <[email protected]>
(cherry picked from commit 626f3b2)

…`.gitignore` entries (#47952)

This was causing our rewriting of the loader's RPATH emulation to fail
after running `make install`, as MSYS2 was rewriting our list that looks
like:

```
../bin/libgcc_s_seh-1.dll:../bin/libopenlibm.dll:@../bin/libjulia-internal.dll:@../bin/libjulia-codegen.dll:
```

Into one that looked like:
```
..\bin\libgcc_s_seh-1.dll;..\bin\libopenlibm.dll;@..\bin\libjulia-internal.dll;@..\bin\libjulia-codegen.dll;
```

By exporting `MSYS2_ARG_CONV_EXCL='*'` for all `stringreplace`
invocations, we dodge this issue, as documented by MSYS2 [0].

[0] https://www.msys2.org/wiki/Porting/#filesystem-namespaces

Tragically, I believe MSYS2 is messing with options such as
`/VERYSILENT` turning them instead into `C:\msys2\VERYSILENT` or
similar.

* Makefile: MSYS2: close path conversion for `DEP_LIBS`

Automatic path conversion will replace `:` with `;`

* Makefile: MSYS2: use `cygpath` for path convert

ref: https://www.msys2.org/docs/filesystem-paths/#manual-unix-windows-path-conversion

* devdoc/win/msys2: add build steps

* devdoc/win/msys2: Add x86/x64 build notes

* devdoc/win/msys2: apply sugestions

Co-Authored-By: Elliot Saba <[email protected]>

* Instead of `CC_WITH_ENV`, scope environment variables to targets

* Fix whitespace

Co-authored-by: Elliot Saba <[email protected]>

This allows our `llvm-config.exe` to find its libraries, which is
necessary on Windows since there's no `RPATH` support.

This was backported as a small piece of 53603f6

…7968)

On later versions of Julia, we have the ability to set CPUs for our
subprocesses more flexibly, but on v1.6 we don't, so let's ignore these
tests for now.

X-ref: 32b1305

Our new CI environment defines environment variables that break this
test.  We have better behaviors on newer Julia versions.

(cherry picked from commit c46834b)

* Generalize Bool parse method to AbstractString

Fixes JuliaStrings/InlineStrings.jl#57.

We currently have a specialization for `parse(Bool, ::Union{String, SubString{String})`
where `true` and `false` are parsed appropriately. The restriction to
`Union{String, SubString{String}}`, however, means we don't get this behavior
for other `AbstractString`s. In the linked issue above, for InlineStrings, we end up
going through the generic integer parsing codepath which results in an `InexactError`
when we try to do `Bool(10)`.

The proposal in this PR takes advantage of the fact that there is only the 2 comparisons
where we do `_memcmp` that require the input string to be "dense" (in memory), and otherwise,
we just do a comparison against a `SubString` of the input string.

Relatedly, I've wanted to introduce the concept of an abstrac type like:

```julia
abstract type MemoryAddressableString <: AbstractString
```

where the additional required interface would be being able to call
`pointer(::MemoryAddressableString)`, since a lot of our string algorithms depend on doing
these kind of pointer operations and hence makes it quite a pain to implement your own
custom string type.

* Apply suggestions from code review

Co-authored-by: Stefan Karpinski <[email protected]>
Co-authored-by: Nick Robinson <[email protected]>

Co-authored-by: Stefan Karpinski <[email protected]>
Co-authored-by: Nick Robinson <[email protected]>
(cherry picked from commit 63830a6)

(cherry picked from commit f6b5157)

(cherry picked from commit 6d678fe)

The dynamic linker needs to unify `llvm::Any::TypeId` across DSOs. In our case
`libjulia-codegen` and `libLLVM`.

See https://github.com/llvm/llvm-project/blob/2bc4c3e920ee078ef2879b00c40440e0867f0b9e/llvm/include/llvm/ADT/Any.h#L30

Fixes: #49121
(cherry picked from commit d8fa3c8)

(cherry picked from commit 1eee6ef)

(cherry picked from commit 432f300)

Fix #43129

(cherry picked from commit 832e46d)