v2.4

What's new?

• Graceful LFS_ASSERT (#514)

  Thanks to @mon, @davidefer, and everyone else who provided feedback, littlefs now supports overriding LFS_ASSERT with a macro that can return an error instead of halting.

  This doesn't change the behavior LFS_ASSERT itself, but just means littlefs only uses LFS_ASSERT in functions that return a signed error code. This is an internal rule littlefs will follow, and thanks to @mon we now have a CI job to prevent breaking this behavior. Outside of littlefs, you're free to use LFS_ASSERT in non-erroring functions, you'll just lose the ability to override LFS_ASSERT in this way.

• New metadata_max config (#502)

  A great idea from @mon, metadata_max provides a workaround for performance issues littlefs faces with large block sizes (#203).

  metadata_max reduces the performance hit of metadata compaction by artificially limiting how much of a metadata block we use. This improves performance since we there is less metadata to compact, but does mean we waste whatever space in the block we don't use. It's interesting tradeoff since the devices where block sizes are large enough for this to be an issue (NAND) also have enough storage to throw at this problem. It's not the best solution, but may make it possible to get past some of these performance bottlenecks.

  @mon notes a 10x speedup in directory traversal/file deletion and a 2x-4x speed up in file reads on a 128KiB NAND block device (MT29F) with metadata_max set to 4KiB.

• Adopt GitHub Actions for CI, and a number of script/Makefile improvements (#516)

  Since this is primarily for development and doesn't affect most users, I will just point to the PR notes in #516.

  If everything works, you will probably already see the new code size/coverage measurements automatically appended to these release notes.

  One of the more interesting things to note is the addition of a post-release process to automatically create PRs against dependent repos. This should help keep dependents such as littlefs-fuse no longer constantly-out-of-date. If you want to add a repo to receive automatic PRs feel free to create a PR against the littlefs.post-release repo.

• Thanks to @embeddedt, lfs_file_seek has a new optimization to avoid flushing caches when seeking to the same position (#511)

• littlefs now performs a sanity check in lfs_format after compacting the superblock, instead of before (#508)

Changes

Code size	Code size (readonly)	Code size (threadsafe)	Code size (migrate)	Code size (error-asserts)	Coverage
14600 B	5586 B	15422 B	16296 B	15212 B	94.7% of 1758 lines

3d4e4f2 Bumped minor version to v2.4
c0cc0a4 Enabled overriding of LFS_ASSERT/TRACE/DEBUG/etc
c9eed1f Add test to ensure asserts can return
e7e4b35 lfs_fs_preporphans ret int for graceful LFS_ASSERT
21488d9 Fixed incorrect documentation in test.py
10a0883 Moved lfs_mdir_isopen behind LFS_NO_ASSERT
47d6b2f Removed unnecessary truncate condition thanks to new seek optimization
745d98c Fixed lfs_file_truncate issue where internal state may not be flushed
3216b07 Use lfs_file_rawsize to calculate LFS_SEEK_END position
6592719 Removed .travis.yml
c911061 Added post-release script, cleaned up workflows
104d651 Reduced build sources to just the core littlefs
6d3e4ac Brought over the release workflow
9d65460 Fixed a recompilation issue in CI, tweaked coverage.py a bit more
b84fb6b Added BUILDDIR, a bit of script reworking
887f366 Switched to lcov for coverage collection, greatly simplified coverage.py
eeeceb9 Added coverage.py, and optional coverage info to test.py
b2235e9 Added GitHub workflows to run tests
6bb4043 Skip flushing file if lfs_file_rawseek() doesn't change position
2b80453 Moved sanity check in lfs_format after compaction
d804c2d Added scripts/code_size.py, for more in-depth code-size reporting
37f4de2 Remove inline_files_max and lfs_t entry for metadata_max
6b16daf Add metadata_max and inline_file_max to config