We have `typeof(Union{}) == Type{Union{}}`, but were treating them
differently in the calculation of `isdispatchtuple`. The compiler
expects `isdispatchtuple` to commute with type equality in various
places, so try to make this consistent.

Fixes #45347

(cherry picked from commit a37dd16)

(cherry picked from commit a5438f9)

This new build doesn't depend on Kerberos

Co-authored-by: Mosè Giordano <[email protected]>
(cherry picked from commit 28d9326)

…e inlining"

This reverts commit 8bca2f4.

…#45534)

(cherry picked from commit 21e5a26)

(cherry picked from commit 6ea9034)

(cherry picked from commit 6f8860c)

(cherry picked from commit 0a55a8e)

* diagonal*offset matrix should throw

(cherry picked from commit 172bddc)

Since the `data` argument possibly does not come from the current
configuration, we should not use `nthreads` here either to filter out
possibly significant data.

(cherry picked from commit b46c14e)

…5615)

`is_foldable` should be more aligned with `Base.@assume_effects`.
This should be backported to 1.8 also.

This reverts commit eb82f18.

I've omitted `invpcid` feature since it's clearly omitted for other CPU's.

Fixes #45657

(cherry picked from commit 3f047f7)

This fixes hundreds of invalidations from Unitful.

Co-authored by: Shuhei Kadowaki <[email protected]>

`@time` now shows if any of the compilation time was spent recompiling invalidated methods. The new percentage is % of the compilation time, not the total execution time.

(cherry picked from commit 7074f04)

(cherry picked from commit 9320fba)

This is theoretically okay, but unlikely to be intended ever.

(cherry picked from commit d81724a)

Does not explicitly close issue #45704, as perhaps the deserialized
module should still be valid after the replacement warning.

(cherry picked from commit ad8893b)

Ensures we do not get easily wedged into bad states.

(cherry picked from commit d0e28af)

Regressions introduced by #45607

(cherry picked from commit 239a1f2)

(cherry picked from commit b49a1b4)

fix #44705

(cherry picked from commit 4115686)

…func` (#44896)

Follows up #44725.

(cherry picked from commit 0deb326)

(cherry picked from commit 4f1c68e)

(cherry picked from commit 9b83dd8)

…ation (#45746)

* Bump LBT to 5.1.1
* Use DESTDIR to install into
* Update LBT checksums

(cherry picked from commit 1d78265)

(cherry picked from commit 1f99ee9)

When egal objects contain identical typevars with different
environments, the resulting subtyping might not be so obvious.

Fix #45703

(cherry picked from commit 8ccf2d6)

This prevents a compiler warning on some systems

(cherry picked from commit 05eb153)

* Bug fix for `max_collect_interval` computation (#45727)

Currently constrained to `totalmem / ncores / 2` for `_P64` which
results in a very short collect interval when you're running with
a smaller number of threads on a machine with many cores.

Changes this to `totalmem / nthreads / 2` which, for two of our
tests, resulted in 40% and 60% runtime reduction (!!) as well as
GC time reduction from 46% to 10% and 64% to 11%.

* Move GC init after threading init

To allow use of `jl_n_threads` in GC initialization.