* Perform path completion filtering based on expanded path
This fixes issues with the cd and rmdir completers not working
correctly due to os.path.isdir returning false for paths like
'~/example'.
* Avoid running subsequent completers for cd/rmdir
Raise StopIteration in the cd and rmdir completers if nothing is
found. This prevents the later completers like the generic 'path'
completer from running and returning incorrect results.
* fix: handle ast.Index and ast.ExtSlice change while Parsing
related to #4100, #4099, #4068, #4038, #4028, #4005,
* fix: py39 raises when lineno & col not given
TypeError: required field "col_offset" missing from keyword
* fix: tuple inside scbscriptor support for py39
* chore: enable py39 ci checks
they were silently failing before
* test: a backlog test for attributes inside subscriptors
this would require debugging deep into the ply parser methods.
* docs: add news item
* test: skip test_rc_with_modified_path on windows
* Update tests/test_parser.py
Co-authored-by: Gil Forsyth <gforsyth@users.noreply.github.com>
* test: mark failing tests for py39 parser
Co-authored-by: Gil Forsyth <gforsyth@users.noreply.github.com>
* add "disown" internal command
This command will remove jobs from the shell's job table, allowing the command
to continue running after the shell exits. Configuration options are provided
to manage how suspended background jobs are handled after being disowned.
* job-ctrl-disown code style update
* job-ctrl-disown style updates + change arg --cont to --continue
* Trim extra "=" in doc for "disown" cmd
Co-authored-by: Gil Forsyth <gforsyth@users.noreply.github.com>
Co-authored-by: Gil Forsyth <gforsyth@users.noreply.github.com>
* Add $HISTCONTROL=ignorespace support
* Add news and fix formatting
* Move ignore space to happend in append and fix test bugs
* invert conditional
* remove leftover debug info
* Fix libc detection on FreeBSD
On FreeBSD /usr/lib/libc.so is an LDScript text file, ctypes.CDLL can't handle
those. So we hardcode the current version number, lest we'd need to use the less
efficient ctypes.util.find_library function. This should be alright, since the
last increase in libc version number on FreeBSD was over 10 years ago.
* sysctlbyname needs bytes, not unicode strs
* Add news item for FreeBSD libc fix
* aliases: Add merging operation to the aliases dict
* tests: Add tests to the aliases merging operation
* news: Add summary about the Alias merging functionality
* chore: bring back codcoverage report
* chore: the run-tests.xsh script can't print to stdout
* test: combine coverage on multiple runs
* test: sort coverage report by coverage percent
* chore (ci): split test step
* docs: add news item for code-coverage
* chore: try coverage parallel
but this will not work with --append mode.
* ulimit: Change default to access soft limits
It makes more sense to manipulate "soft" limits by default, because
those are the limits intended to be user-adjustable. This is also
what bash does. Originally, I made ulimit to default to "hard" limits,
because I was for some reason convinced that it's what bash does.
* Changelog entry
* refactor: rewrite xontribs/jedi.xsh -> xontribs/jedi.py to take advantage of python tooling
* chore: update elm ci workflow
* test: update test for jedi xontrib
this will further speedup the workflow.
- also the macos and windows should now have proper cache with conda
- use jinja to template workflows that looks the same
* Support path completion from path literal
* Test path completion when prefix is partial p-string
* Update news
* Update news
* fix test: expect string literal
* Support completion of raw-path strings
* Resolve raw-string issues when testing on Windows
* Simplify p-string prefix checks
* Fix test
* Update news
* fix news: rst
* Use raw path string to simplify test
* Test path from partial string
* Try case sensitive completion
