Endianess code should use memcpy instead of unions. This apparently
is more correct on old, optimizing compilers. Technically, this is
compilers being really stupid but we work with what we got.
That said, this endianess code is more complicated than needed.
The backing buffer for weak arrays and tables was not freed upon
being garbage collected. This shows up in traces and valgrind. Verified
by running `make valtest` with changes.
2 issues:
- With poll backend, we were polling for writes even after we finished
writing. Presents as wasting a lot of CPU.
- Fixes strange closing behavior of chat server.
Weak containers did not preserve their weakness when marshalled. This
fixes that for tables and arrays, as well as adds some tests for this.
Also exposes functions for creating weak tables in janet.h
Win32 abstraction will use ReadDirectoryChanges with overlapped
IO to get results. Some work will be required to allow for single file
watches, as well as allow for recursive watching on Linux.
Unfortunately, various operating systems have very differnet
abstractions here. I would rather expose inotify, kqueue, and
ReadDirectoryChanges fairly transparently before
adding shims to make cross platform code easier.
Number literals can now take an optional "representation" suffix
- Use `:n` for normal numbers (IEEE-754 doubles)
- Use `:s` for signed 64 bit integers
- Use `:u` for unsigned 64 bit integers
- Other suffix will fallthrough the usual parseing logic. This means
that they will only possibly resolve to symbols if they start with -,
+, or .
The syntax does not collide with any existing valid Janet and is only
enabled with JANET_INTTYPES. This also leaves open a syntax for other
number types such as bignums, ratios, decimals, etc.
Establishing a convention for scripts is beneficial for various tools.
However, we do not install scripts on anyones PATH - instead they go
to a self contained (dyn *syspath*) /bin folder which could be added to
path, or symlinks could be added.
Ctrl + left/right arrow would simply input "5D"/"5C" into the REPL
which was useless and confusing
With this change, it instead goes to the previous/next word which is
usually expected in readline-like interfaces
