This PR adds support for
[mp3rgain](https://github.com/M-Igashi/mp3rgain) as an alternative to
mp3gain/aacgain in the ReplayGain plugin's command backend.
## What is mp3rgain?
mp3rgain is a modern Rust rewrite of mp3gain that provides:
- **CLI-compatible drop-in replacement** for mp3gain - same command-line
interface and output format
- **Support for both MP3 and AAC/M4A formats** (like aacgain)
- **Security fixes** for vulnerabilities in the original mp3gain:
- CVE-2021-34085 (Critical, CVSS 9.8) - buffer over-read in
`WriteMP3GainAPETag`
- CVE-2019-18359 (Medium, CVSS 5.5) - buffer over-read in
`ReadMP3APETag`
- **Memory-safe implementation** in Rust
- **Works on modern systems** including Windows 11 and macOS with Apple
Silicon
- **Active maintenance** (original mp3gain has not been updated since
2009)
## Changes
1. **beetsplug/replaygain.py**:
- Add `mp3rgain` to the command search list (prioritized first since
it's more secure)
- Update `format_supported()` to correctly handle mp3rgain's AAC/M4A
support
- Update error message to mention mp3rgain
2. **docs/plugins/replaygain.rst**:
- Add mp3rgain to the backend documentation
- Include installation instructions for various platforms
- Document the security benefits
## Testing
mp3rgain has been tested to be fully compatible with the existing
command backend:
- Supports all the same command-line flags (`-o`, `-s s`, `-k`, `-c`,
`-d`, etc.)
- Produces identical tab-delimited output format
- Returns the same exit codes
## Why prioritize mp3rgain in the search order?
The original mp3gain has known security vulnerabilities that have not
been fixed (the project has been inactive since 2009). By searching for
mp3rgain first, users who have it installed will automatically benefit
from the security improvements, while maintaining full backward
compatibility for users who only have mp3gain/aacgain installed.
## Description
was getting permission error because after png is converted to jpg beets
want to delete the png but somehow it is still being used causing the
import to fail. this temporarily fixes the import but still needs a
proper way to know what is using the file and how to delete it.
```
File "C:\Users\DELL\projects\_myForks\beets\beetsplug\fetchart.py", line 1321, in fetch_art
candidate = self.art_for_album(task.album, task.paths, local)
File "C:\Users\DELL\projects\_myForks\beets\beetsplug\fetchart.py", line 1413, in art_for_album
out.resize(self)
~~~~~~~~~~^^^^^^
File "C:\Users\DELL\projects\_myForks\beets\beetsplug\fetchart.py", line 218, in resize
self._resize(plugin, current_check)
~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\DELL\projects\_myForks\beets\beetsplug\fetchart.py", line 246, in _resize
self.path = ArtResizer.shared.reformat(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
self.path,
^^^^^^^^^^
plugin.cover_format,
^^^^^^^^^^^^^^^^^^^^
deinterlaced=plugin.deinterlace,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "C:\Users\DELL\projects\_myForks\beets\beets\util\artresizer.py", line 658, in reformat
os.unlink(path_in)
~~~~~~~~~^^^^^^^^^
PermissionError: [WinError 32] The process cannot access the file because it is being used by another process: b'C:\\Users\\DELL\\AppData\\Local\\Temp\\beets\\beetsplug_fetchart\\4kqx2um2.png'
```
when importing
https://musicbrainz.org/release/5744ddb7-e9b6-4b46-a55c-38e75aa95460
beet config
```yaml
fetchart:
minwidth: 500
maxwidth: 3000
max_filesize: 3500000
sources:
- coverart: release
- coverart: releasegroup
- itunes
- amazon
- filesystem
- albumart
- '*'
cautious: yes
cover_names: cover front art artwork folder album
store_source: yes
cover_format: JPEG
auto: yes
quality: 0
enforce_ratio: no
high_resolution: no
deinterlace: no
```
- Add Tool type alias and SUPPORTED_FORMATS_BY_TOOL class variable
- Refactor __init__ to use shutil.which() and set cmd_name early
- Simplify format_supported() to use dictionary lookup
Given that @frigginbrownie's #5556 PR received some thumb ups but they
haven't responded since a while ago, I'm creating this PR to merge their
fix (I could not commit in the PR branch since I have no permissions to
push to their fork).
Supersedes: #5556
Copying @frigginbrownie description from #5556:
According to the docs, the auto_keep function will "Convert your files
automatically on import to dest but import the non transcoded version."
This is true but not 100% accurate. In cases where no conversion is
required (say, importing lossy files where there's no need to convert),
auto_keep will copy the files to dest.
This behavior results in duplicate files being created on import when
the auto_keep function is set to yes - a lossy file will be imported
into the default directory (say /music) and then copied to the dest
location (say /transcodes).
This is ideal if you wish to have all music formats in your default
directory (lossy and lossless) and all lossy files (original imports and
transcodes) in a secondary directory (say /lossy).
But what if you want a separate directory of all music you've
transcoded? auto_keep won't provide that, as it copies lossy files to
the dest location. In addition, if the dest is set to the same location
as default directory, auto_keep will copy lossy files into the same
directory that beets previously imported files into, resulting in the
directory having two files for each file in an album. If you use paths
(say to have singletons imported into /music/singles), auto_keep will
import the file into the path location, then copy the file to the dest,
creating directories to match the path.
Unlike with the auto option or using "beet convert", auto_keep does not
follow the never_convert_lossy_files or no_convert options and will not
validate whether files need to be converted or copied on import to dest
- it transcodes or it copies, no questions asked.
This change updates the auto_convert_keep function to filter items using
should_transcode. This way, if the user sets never_convert_lossy_files
to no or no_convert: 'format:mp3', lossy files will not be copied to the
dest, while lossless files will be converted to the dest (perfect for a
seperate /transcodes directory). If the user sets
never_convert_lossy_files to yes, lossy files will to be copied to the
dest and lossless files will be converted to the dest (perfect for a
/lossy directory). In turn, this change makes behavior consistent with
"beet convert" and the auto option.
Fixes#3706.
### Issue
Comment tags are written to file even if option 'from_scratch' is used.
The same tags are not written to the file if imported together with
other files as album. Therefore 'from_scratch' is not working as
described in the documentation.
### Solution
1. Add test: Adapt the function from the 'regular' import class and
insert it in the class for the singleton import test.
2. Fix bug : Add check for 'from_scratch' option. If used, clear
metadata before applying 'new' metadata with autotag.
3. No documentation change needed. Option now works as described in the
documentation.
4. Add changelog.
mp3rgain is a modern Rust rewrite of mp3gain that provides:
- CLI-compatible drop-in replacement for mp3gain
- Support for both MP3 and AAC/M4A formats (like aacgain)
- Fixes for CVE-2021-34085 (Critical, CVSS 9.8) and CVE-2019-18359 (Medium)
- Memory-safe implementation in Rust
- Works on modern systems (Windows 11, macOS Apple Silicon)
Changes:
- Add mp3rgain to the command search list (prioritized first)
- Update format_supported() with more robust command name detection
using os.path.basename() and startswith() instead of substring matching
- Update documentation with installation instructions
See: https://github.com/M-Igashi/mp3rgain
## Summary
This PR updates typing and linting across the codebase and enables
stricter `ruff` checks for Python 3.10:
1. Enable `tool.ruff.lint.future-annotations`
Very handy feature released in `0.13.0`: if required, it _automatically_
adds `from __future__ import annotations` and moves relevant imports
under `if TYPE_CHECKING`:
```py
# before (runtime import)
from beets.library import Library
# after
from __future__ import annotations
from typing import TYPE_CHECKING
if TYPE_CHECKING:
from beets.library import Library
```
2. Set `tool.ruff.target-version = "py310"`
This enforced PEP 604 unions in the codebase:
```py
# before
SQLiteType = Union[str, bytes, float, int, memoryview, None]
# after
SQLiteType = str | bytes | float | int | memoryview | None
```
3. Enable `RUF*` family of checks
- Remove unused `# noqa`s
- Ignore unused unpacked variables
```py
# before
likelies, consensus = util.get_most_common_tags(self.items)
# after
likelies, _ = util.get_most_common_tags(self.items)
```
- Avoid list materialization
```py
# before
for part in parts + [","]:
# after
for part in [*parts, ","]:
```
- And, most importantly, **RUF012**: use `ClassVar` for mutable class
attributes
- This underlined our messy `BeetsPlugin.template_*` attributes design,
where I have now defined `BeetsPluginMeta` to make a clear distinction
between class and instance attributes. @semohr and @asardaes I saw you
had a discussion regarding these earlier - we will need to revisit this
at some point to sort it out for good.
- It also revealed a legitimate issue in `metasync.MetaSource` where
`item_types` were initialised as an instance attribute (but luckily
never used).
The beets-usertag plugin was originally created by igordetigor, but has
been unmaintained for some time now. Recently, I decided to add some
features to it and, with Ingo's blessing, publish on PyPI. He's also
okay with me replacing the link in the plugins list with my fork.
Looks like our hard-coded API credentials expired. This PR removes
expired credentials and makes it clear to the user that they must
provide their credentials.
Fixes#6270
