rmlint – Reference

● cpx May 26, 2026 4 min read System Tools

Version: 2.10.x (Ludicrous Lemur)

Purpose: Find duplicate files, empty dirs, broken links and other space waste

Key principle: rmlint never deletes anything — it produces scripts that do

Table of Contents

3 Minutes Read

Syntax

rmlint [PATHS...] [//] [TAGGED_PATHS...] [-] [OPTIONS]

Paths before // are untagged (default). Paths after // are tagged (preferred as originals).

Lint Types (`-T` / `--types`)

Short	Long	Description
`df`	`duplicates`	Duplicate files
`dd`	`duplicatedirs`	Duplicate directories (same as `-D`)
`ef`	`emptyfiles`	Empty files
`ed`	`emptydirs`	Empty directories
`bl`	`badlinks`	Broken symbolic links
`bi`	`badids`	Files with bad UID/GID
`ns`	`nonstripped`	Binaries with debug symbols (must enable explicitly)

Groups: all · defaults (all minus ns) · minimal (defaults minus ef,ed) · minimaldirs · none

General Options

Short	Long	Default	Description
`-T`	`--types`	`defaults`	Lint types to search for
`-o`	`--output`	sh + pretty + summary + json	Set output formatters (overrides defaults)
`-O`	`--add-output`	—	Add output formatter (preserves defaults)
`-c`	`--config`	—	Configure a formatter (`-c sh:link`)
`-a`	`--algorithm`	`blake2b`	Hash algorithm for dupe detection
`-p`	`--paranoid`	—	Byte-by-byte comparison
`-P`	`--less-paranoid`	—	Use faster/weaker hash
`-v`	`--loud`	—	Increase verbosity (stackable)
`-V`	`--quiet`	—	Decrease verbosity (stackable)
`-g`	`--progress`	off	Show progress bar with sane defaults
`-D`	`--merge-directories`	off	Find duplicate directory trees
`-j`	`--honour-dir-layout`	off	Require same path layout for dupe dirs
`-y`	`--sort-by`	none	Sort output groups
`-w`	`--with-color`	on	Colored output
`-z`	`--perms`	no check	Only consider files with given permissions (`rwx`)

Hash Algorithms (`-a`)

Bits	Algorithms
512	`blake2b` · `blake2bp` · `sha3-512` · `sha512`
384	`sha3-384`
256	`blake2s` · `blake2sp` · `sha3-256` · `sha256` · `highway256` · `metro256` · `metrocrc256`
160	`sha1`
128	`md5` · `murmur` · `metro` · `metrocrc`
64	`highway64` · `xxhash` (not recommended)

Paranoia shortcuts: -p = byte-by-byte · -P = highway256 · -PP = metro256 · -PPP = metro

Traversal Options

Short	Long	Default	Description
`-s`	`--size`	`1` (≥1 byte)	Size range filter (`100KB-2M`)
`-d`	`--max-depth`	INF	Max recursion depth
`-l`	`--hardlinked`	default	Treat hardlinks as dupes
`-L`	`--no-hardlinked`	—	Ignore all but one of hardlinked set
`-f`	`--followlinks`	—	Follow symlinks
`-F`	`--no-followlinks`	—	Ignore symlinks completely
`-@`	`--see-symlinks`	default	Treat symlinks as small files
`-x`	`--no-crossdev`	—	Stay on same filesystem
`-X`	`--crossdev`	default	Allow crossing mountpoints
`-r`	`--hidden`	—	Include hidden files/dirs
`-R`	`--no-hidden`	default	Skip hidden files/dirs
`-b`	`--match-basename`	—	Only match files with same basename
`-B`	`--unmatched-basename`	—	Only match files with different basename
`-e`	`--match-extension`	—	Only match files with same extension
`-i`	`--match-without-extension`	—	Match by basename minus extension
`-n`	`--newer-than-stamp`	—	Filter by mtime from stamp file
`-N`	`--newer-than`	—	Filter by mtime (ISO8601 or epoch)

Original Detection Options

Short	Long	Description
`-k`	`--keep-all-tagged`	Never delete files in tagged (`//`) paths
`-K`	`--keep-all-untagged`	Never delete files in untagged paths
`-m`	`--must-match-tagged`	Only find dupes with ≥1 file in tagged paths
`-M`	`--must-match-untagged`	Only find dupes with ≥1 file in untagged paths
`-S`	`--rank-by`	Sort criteria for original selection (default: `pOma`)

Rank-By Keys (`-S`)

Key	Keep…	KEY	Keep…
`m`	oldest mtime	`M`	newest mtime
`a`	first alphabetically	`A`	last alphabetically
`p`	first named path	`P`	last named path
`d`	lowest depth	`D`	highest depth
`l`	shortest basename	`L`	longest basename
`h`	lowest hardlink count	`H`	highest hardlink count
`o`	lowest external hardlinks	`O`	highest external hardlinks
`r`	paths matching regex	`R`	paths not matching regex
`x`	basenames matching regex	`X`	basenames not matching regex

Default pOma: prefer first-named path → most external hardlinks → oldest → alphabetical

Caching Options

Flag	Description
`--replay`	Re-read a previous `.json` run (no disk I/O)
`-C` / `--xattr`	Shortcut for `--xattr-read --xattr-write --write-unfinished`
`--xattr-read`	Read cached checksums from extended attributes
`--xattr-write`	Write checksums to extended attributes after hashing
`--xattr-clear`	Clear cached checksums from extended attributes
`-U` / `--write-unfinished`	Include partially-hashed files in output

Miscellaneous Options

Short	Long	Default	Description
`-t`	`--threads`	`16`	Threads for traversal & hashing
`-u`	`--limit-mem`	—	Max memory for hashing (`-u 512M`)
`-q`	`--clamp-low`	`0`	Read file from this offset
`-Q`	`--clamp-top`	`1.0`	Read file up to this offset
`-Z`	`--mtime-window`	`-1`	Max mtime difference (seconds) for dupes

Output Formatters (`-o` / `-O`)

Formatter	Description
`sh`	Shell script with removal commands (default)
`json`	JSON document (streamable) (default)
`csv`	Comma-separated values
`py`	Python script + JSON
`pretty`	Colored real-time display (default)
`summary`	Post-run statistics (default)
`progressbar`	Progress bar (use with `-g`)
`fdupes`	fdupes-compatible output
`stamp`	Timestamp for `--newer-than-stamp`
`uniques`	One unique file path per line

Shell Script Handlers (`-c sh:handler=...`)

Handler	Description
`clone`	FIDEDUPERANGE ioctl (btrfs, kernel 4.2+)
`reflink`	Reflink dupe → original (CoW filesystems)
`hardlink`	Hardlink dupe → original (same partition)
`symlink`	Symlink dupe → original (always works)
`remove`	`rm -rf` (default)
`usercmd`	Run user command from `-c sh:cmd='...'`

Shortcuts: -c sh:link = clone,reflink,hardlink,symlink · -c sh:hardlink = hardlink,symlink · -c sh:symlink = symlink

Standalone Commands

Command	Description
`rmlint --gui`	Launch Shredder GUI
`rmlint --equal <p1> <p2>`	Test if two paths have identical content
`rmlint --dedupe <src> <dest>`	Dedupe via reflink (CoW filesystems)
`rmlint --is-reflink <f1> <f2>`	Test if two files share physical storage

Common Recipes

# Basic scan with progress bar
rmlint -g ~/data

# Find dupes, prefer originals in master dir
rmlint ~/backup // ~/master -k -m

# Deduplicate with hardlinks (fallback to symlinks)
rmlint -c sh:hardlink ~/photos

# Deduplicate on btrfs
rmlint -c sh:link ~/snapshots

# Only files > 1MB, byte-by-byte comparison
rmlint -p -s 1M ~/documents

# Find duplicate directory trees
rmlint -D ~/projects

# Replay previous run with different ranking
rmlint --replay rmlint.json ~/data -S MaD

# Use xattr caching for large datasets
rmlint --xattr ~/nas/media

# Custom command on duplicates
rmlint -o sh -c sh:cmd='echo "dupe: $1 orig: $2"' ~/data

# Only broken links, empty dirs/files
rmlint -T "bl,ed,ef" ~/home

# Filter by mtime
rmlint -N 2025-01-01T00:00:00 ~/downloads

Source: rmlint man page · rmlint docs · GitHub

https://www.linkedin.com/posts/lewiswalkerai_the-layers-of-an-agentic-ai-platform-according-share-7448389610281418752-INpU?utm_source=share&utm_medium=member_ios&rcm=ACoAAAIKz-8B5JvHFavurAchVmkTrGphZWE9CW4

0 0 votes

Article Rating

0 Comments

Oldest

Newest Most Voted

rmlint – Reference

Syntax

Lint Types (-T / --types)

General Options

Hash Algorithms (-a)

Traversal Options

Original Detection Options

Rank-By Keys (-S)

Caching Options

Miscellaneous Options

Output Formatters (-o / -O)

Shell Script Handlers (-c sh:handler=...)