This commit merges changes that were developed in the "racket7" repo.
See that repo (which is no longer modified) for a more fine-grained
change history.
The commit includes experimental support for running Racket on Chez
Scheme, but that "CS" variant is not built by default.
Add a `_bytes` variant type that will work more consistently with
Racket-on-Chez, where the representation of a byte string does not
include an implicit nul terminator.
Faster for the case that an impersonator has a small number of
impersonator properties. To enable hand-coded search, a small number
of properties are now kept in a vector instead of a hash tree.
Commit 89512eda had accientally disabled the JIT's fast
path for structure-type property predicates and selectors.
Re-enable it, but but repair a problem with the way that
impersonator-property operations are handled by the same
code.
Before 544b7a3d53, an incorrect barrier check prevented applying
a continuation captured in other threads from being applied in a
thread created with `call-in-nested-thread`. There's no good
reason to prohibit such an application, and now that the bogus
check is gone, an explicit prompt is needed in the implementation
to make it behave correctly.
Provide unsafe functions for working with file descriptors and
sockets. Although more functions are potentially useful, these
reflect the one scurrently exported by the C API.
When a composable continuation has continuation marks that should
be merged with marks in the immediate continuation frame when
the continuation is applied, then mergeing did not always work.
It only worked in the case that the merge cadidates are the only
marks, because the merging check used the wrong end of the captured
sequence of marks.
Using `(local-expand <expr> 'module-begin <stops> def-ctx)` didn't
work right, because definitions added to `def-ctx` were not visible.
(While adding definitions before `module-begin` expansion is an unusual
thing to do, there's no reason that it has to fail.)
Repairs a problem with ce9894c8bf, where a large "inlined" vector
allocation is not actually inlined, but other parts of the JIT
assume that it will behave as inlined --- which implies that the
runstack will be left unchanged after the call.
Closes#1868
Commit 00d438cfbe made an attempt at this,
but this commit does it in a much more careful way, based on manually
emulating how the macroexpander expands module* forms in order to allow
splicing-syntax-parameterize to apply even within #%module-begin forms
introduced by the expander.
Normally, it's impossible to generate lots of `eq?`-hashing
collisions, but when the compiler inlines a function, it can duplicate
variables in a way that gives each copy the same `eq?` hash code. The
immutable-hash tree implementation failed when more than 32 collisions
occurred (which triggers a subtree in the collision node).
It's similarly very difficult to generate > 32 values that collide on
`eqv?` hashes but are not `eqv?` (although it must be possible using
exact rationals or complex numbers).
