https://github.com/brotzeit/rustic

Update core-test

Typos found with codespell.

Closes #198

Fix doc

#214

Use elisp core by default

* Add empty CHANGELOG.md

* Generate CHANGELOG template

```python
#!/usr/bin/env python3

import subprocess

def _get_tag_date(tag):
    git_show = subprocess.run(["git", "show", tag, "--format=%cd", "--date=short", "-s"], capture_output=True)
    return git_show.stdout.decode("utf-8").strip()

def main():
    git_tag = subprocess.run(["git",  "tag"], capture_output=True)
    versions = list(reversed([e[1:] for e in git_tag.stdout.decode("utf-8").split("\n") if e]))
    for version in versions:
        print(f"## [{version}] - {_get_tag_date('v' + version)}")
        print()
        print()

    print(f"[Unreleased]: https://github.com/editorconfig/editorconfig-emacs/compare/v{versions[0]}...HEAD")
    for version, version_prev in zip(versions, versions[1:]):
        print(f"[{version}]: https://github.com/editorconfig/editorconfig-emacs/compare/v{version_prev}...v{version}")
    print(f"[{versions[-1]}]: https://github.com/editorconfig/editorconfig-emacs/releases/tag/v{versions[-1]}")
    return

main()
```

* Paste release notes from github releases

* Add release note for unreleased

* Fix line

Previously, `editorconfig-get-properties` could return either an equal-based
hash table (if the `editorconfig-exec-path` executable was run) or an eql-based
hash table (if the internal elisp implementation was used).  Since the former is
keyed on symbols anyway, it makes sense to make them both use eql for
consistency.

According to [the editorconfig specification][1],

> For any pair, a value of `unset` removes the effect of that pair, even if it
> has been set before. For example, add `indent_size = unset` to undefine the
> `indent_size` pair (and use editor defaults).

Yet, the test suite for editorconfig *cores* [indicates][2] that an editorconfig
shouldn't ignore values of "unset", and should return them as-is.

Therefore it's up to editorconfig *plugins* to discard values of "unset" at the
point of use.

Before this commit, the only place in the code where the value "unset" was
checked for was in `editorconfig-set-major-mode-from-ext`, whereas in other
places, such as `editorconfig-set-indentation`, it was not checked.

After this commit, values of "unset" are immediately removed from the properties
hashtable before it is used to do anything in `editorconfig-apply`.  This
ensures that properties with value "unset" will be treated indistinguishably
from unspecified properties.

This fixes the behavior shown in this gist:

  https://gist.github.com/kini/560673dbff4b8e42bb40a82a8ca6e5b1

[1]: https://editorconfig-specification.readthedocs.io/en/latest/#supported-pairs
[2]: editorconfig/editorconfig-core-test#12

Note that `indent-tabs-mode` and `tab-width` are both "buffer-local whenever
set" variables so there's no need to run `make-local-variable` on them, as I
understand it.

Always handle "unset" values as non-present

* Update README to help bootstrap users

* Update doc

* Fix doc

* Update

* Updat

Update doc

* Fix issue references in CHANGELOG

* Fix commit reference link

* Add Emacs 27.1 to travis test target

* Add GitHub Actions build.yaml

* Fix evm 27.1 version

* Remove github actions build.yaml

TRAMP paths looks like `/method:user@host:/path/to/file`, per https://www.gnu.org/software/tramp/#Quick-Start-Guide .

editorconfig via Tramp over SSH causes find-file to lag, as it traverses up the remote file hierarchy, each step incurring a network roundtrip.

Co-authored-by: Hong Xu <hong@topbug.net>

Move aliases before their referents

Co-authored-by: 10sr <8.slashes@gmail.com>

Add Github Actions CI

* Update README 2

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Use revert-buffer-with-coding-system to set coding system

* Avoid infinite call

* Do not ask user for revert when buffer is not modified

* Refactor -set-coding-system

* Do not use revert when the file does not exist yet

* Refactor -set-coding-system

Some buffers get confused if you keep setting their mode, one example
is git-rebase-mode. This is an example buffer of a class of
special-mode buffers which are usually (but not always) non-file
buffers. Either way these buffers are usually special purpose
interactive buffers that will only get confused if editorconfig tries
to mess around with them.

To exclude remote files again, add following to the init.el:

    (add-to-list 'editorconfig-exclude-regexps
                 (eval-when-compile
                   (rx string-start "/" (or "http" "https" "ftp" "sftp" "rsync" "ssh") ":")))

* Remove file type experimental feature temporarily

* Do some refactors

* Ignore failures of tests for file type feature

* Make editorconfig-get-properties-function accept one argument

* Implement editorconfig--advice-find-file-noselect

* Define editorconfig-error

* Implement advices

* Update adviced function body

* Update function name

* Refactor exclude functions

* Add more error handlings

* Refactor editorconfig--advice-find-file-noselect

* Add editorconfig-2-mode

* Add nadvice to dependency list

* Use display-warning instead of message for debug log

* Update README

* Fix editorconfig-2-mode for read-only-mode

* Add test for editorconfig-2-mode