| Age | Commit message (Collapse) | Author |
|---|
|
- ksh93 uses headers such as <linux/fs.h> and can use other libcs,
so remove the glibc dependency.
- Add libmd as a dependency (this is used by the cksum and md5sum
builtins accessable via the virtual /opt/ast/bin path or
'builtin' built-in commmand).
- Also make the PGO tmpdir with mktemp(1) so that it's potentially
stored on a RAMDisk.
|
|
It only took two weeks...
|
|
(note: Martijn, I know you *are* busy with real life, and
I can understand that since my life has also been hell
for the past couple years, but get your damn ass moving.
Leaving a vulnerability unpatched in the git repo for
multiple days is ludicrous. Also, *please* put out
a CVE using GitHub's security advisory feature. I don't
have the capability for that because I lack the required
authorization for it, so you need to do it!)
|
|
Dynamically linked ksh93 binaries now default to providing all
libcmd builtins. Subsequently, SHOPT_ALL_LIBCMD is only relevant
for statically compiled binaries, which we don't install, so the
code for enabling it no longer has any effect and can be safely
removed.
For more info, see: https://github.com/ksh93/ksh/commit/f15e2c41
|
|
|
|
|
|
The runtime performance difference is nonexistent, while build times
with tcc doing an extra build are slightly slower.
(In the original PGO gist script this code was supposed to avoid a
bug in pty. That bug appears to have since been fixed.)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Most of the installation is now handled by the bin/package script
itself, rather than being done manually. For this to work the
file(1) command bin/package uses to determine the architecture must be
invoked with the --no-sandbox flag, since fakeroot will ordinarily
prevent it from functioning because of seccomp restrictions. The patch
for this was rejected upstream.
For more information: https://github.com/ksh93/ksh/pull/769
|
|
|
|
See https://github.com/ksh93/ksh/commit/f2bc1f45
|
|
|
|
|
|
If the build flags in makepkg.conf change after the previous build, the
bin/package script will force the build to fail. Delete the build
directory beforehand to prevent this scenario.
|
|
|
|
Set ${_all_libcmd} to '1' or 'yes' to enable the full set of ksh
builtins available in libcmd. These are listed via the 'builtin'
built-in and can be executed via the /opt/ast/bin virtual file path or
by loading them with builtin(1).
This also updates the fallback to use -O2 for better performance
(although in normal circumstances it's never used).
|
|
- The documentation for libast and libshell is now installed to
/usr/share/man/man3ast (to avoid filename conflicts with other
packages).
- The COPYRIGHT file from the repository is now also installed
to the license folder alongside LICENSE.md.
- Reorganized the PKGBUILD with comment updates.
|
|
Just because an entry in /etc/shells exists for /bin/*ksh doesn't mean
an equivalent entry for /usr/bin also exists in the file. This could
cause the following error to occur in some rare situations:
The value for the SHELL variable was not found in the /etc/shells file
A for loop is now used to explicitly check for both entries, which
should mitigate that rare issue.
|
|
sample.kshrc:
- Only run id(1) once and store the result in ${euid}, which avoids
running the id command every time the PS1 prompt is printed. While
this optimization is minor and likely won't have a noticable effect
for everyone, on my own machine this change removes a small (and
annoying) delay after pressing the Enter key at an empty prompt.
- Set .sh.value instead of PS1 in the PS1.get discipline function.
(This change is purely for stylistic reasons.)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Currently, the only actively maintaned fork of ksh93 is ksh93u+m. This
commit updates the ksh93-git AUR package to use 93u+m's dev branch.
Other changes:
- Support for pfksh was removed from ksh93u+m, so drop that from the AUR
package.
- Add support for rksh and ksh93 to /etc/shells.
- Update the example kshrc for use with ksh93u+m (I've made sure that it
remains backward compatible with ksh93u+ and ksh2020, should anyone
happen to use an older version of ksh).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|