* fix: detect compressed kernel modules (.ko.gz, .ko.xz, .ko.zst)
The linux-kernel-cataloger only matched plain *.ko files, missing
compressed modules produced when CONFIG_MODULE_COMPRESS is enabled
(common on Debian 13 / Ubuntu 24.04+). This resulted in near-zero
module packages being reported for such filesystems.
Changes:
- Add *.ko.gz, *.ko.xz, *.ko.zst glob patterns to both the cataloger
and capabilities.yaml so the file resolver picks up compressed modules
- Add decompressedModuleReader() which detects the extension and
transparently decompresses via compress/gzip, ulikunitz/xz, or
klauspost/compress/zstd before handing the ELF bytes to the existing
parseLinuxKernelModuleMetadata parser
- Promote github.com/klauspost/compress from indirect to direct dependency
- Add unit tests covering all three compression formats plus the
uncompressed baseline, using a programmatically generated minimal ELF
Fixes#4721
Signed-off-by: Will Bates <william.bates11@outlook.com>
* address reading archives into memory
Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>
---------
Signed-off-by: Will Bates <william.bates11@outlook.com>
Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>
Co-authored-by: Will Bates <william.bates11@outlook.com>
Co-authored-by: Alex Goodman <wagoodman@users.noreply.github.com>
* Rework release workflow to canonical shape
Replace the custom quality-gate job with the reusable check-version-available
and check-gate workflows from anchore/workflows. Remove the phase
workflow_dispatch input; the install-script-only path is now a standalone
workflow (release-install-script.yaml) that can be triggered independently.
- add version-available and check-gate jobs using pinned anchore/workflows SHA
- remove phase input and quality-gate job
- release job now needs [check-gate, version-available]
- release-install-script job no longer conditionally skips based on phase
- add release-install-script.yaml for standalone install script runs
- set permissions: {} at workflow level (contents pushed to release job)
- add concurrency: group: release
Signed-off-by: wagoodman <wagoodman@gmail.com>
Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>
* Tighten workflow-level permissions to {}
Change top-level permissions from contents: read to {} in validations.yaml
and validate-github-actions.yaml, pushing the needed contents: read down
to each job that performs a checkout.
Signed-off-by: wagoodman <wagoodman@gmail.com>
Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>
* keep install script phase, remove workflow
Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>
* remove schema detection workflow
Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>
---------
Signed-off-by: wagoodman <wagoodman@gmail.com>
Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>
The elixir-binary and elixir-library classifiers' regexes only matched
the bare semver triplet (and a single sub-segment for the library), so
release-candidate elixir images were either missed entirely or had
their version truncated:
$ syft -q elixir:1.12.0-rc | grep elixir # nothing
$ syft -q elixir:1.13.0-rc.0 | grep elixir
elixir 1.13.0 binary # truncated, "-rc.0" lost
Extend the version capture group to optionally include
"-<a-z0-9>+(\\.<digits>)?" so "1.12.0-rc.1", "1.13.0-rc.0", etc. match
exactly as the elixir.app and the binary's ELIXIR_VERSION line have
them.
Add a logical fixture under testdata/classifiers/snippets/elixir/
1.12.0-rc.1/linux-amd64 (cloned from the existing 1.19.1 fixture with
just the version strings changed) and register it in
Test_Cataloger_PositiveCases.
Closes#4819
Signed-off-by: Chris (ChrisJr404) <11917633+ChrisJr404@users.noreply.github.com>
Co-authored-by: Chris (ChrisJr404) <11917633+ChrisJr404@users.noreply.github.com>
A trailing slash on --exclude (e.g. './lib/') is dropped during pattern
normalization but doublestar.Match still requires an exact string match,
so the resulting pattern silently matches nothing and the directory is
not excluded. Strip a trailing slash so './lib/' and './lib' behave the
same.
Fixes#4839
Signed-off-by: ChrisJr404 <chris@hacknow.com>
* fix(debian): only parse machine-readable copyright files with Format header
Only parse debian/copyright files as machine-readable DEP-5 format when
they contain the mandatory Format header field pointing to the copyright
specification URI. Files without this header are free-form text and
should not have License: regex patterns applied to them, which previously
produced nonsensical results like "#", "Permission", "This", "see" for
non-machine-readable files.
The fallback license classifier in the debian cataloger will handle
non-machine-readable files by doing full-text license identification.
Closes#4708
Signed-off-by: Bahtya <bahtya@users.noreply.github.com>
Signed-off-by: Bahtya <bahtayr@gmail.com>
Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>
* decompose parseLicensesFromCopyright to address linting issues
Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>
---------
Signed-off-by: Bahtya <bahtya@users.noreply.github.com>
Signed-off-by: Bahtya <bahtayr@gmail.com>
Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>
Co-authored-by: Bahtya <bahtayr@gmail.com>
Co-authored-by: Alex Goodman <wagoodman@users.noreply.github.com>
Archives may contain directory entries with read-only permissions
(e.g., META-INF/ with mode 0555 in JARs). When extracted to temp
directories, these restricted permissions prevented writing files
into those directories, causing "permission denied" errors for
subsequent file entries like META-INF/MANIFEST.MF.
Always use 0755 for extracted directories since they are temporary
working directories owned by the extraction process.
Signed-off-by: Thomas Bechtold <thomasbechtold@jpberlin.de>
The java-pom-cataloger was processing pom.xml files found inside
META-INF/maven/ directories, which are archive metadata embedded by the
Maven build process. These are verbatim copies of the original build POM
containing unresolved ${property} references and dependency versions
inherited from parent POMs that are unavailable inside the archive. This
caused phantom packages with UNKNOWN versions to appear when recursive
archive cataloging opened JARs and exposed their internal pom.xml files
to the pom cataloger.
The java-archive-cataloger already uses pom.properties (which always
contains resolved values) to identify these archives, so the pom
cataloger should skip META-INF/maven/ pom.xml files entirely.
Signed-off-by: Thomas Bechtold <thomasbechtold@jpberlin.de>
Add integration tests covering the full archive extraction pipeline:
- tar.gz containing Python package metadata (single-level)
- zip containing a JAR with MANIFEST.MF (two-level, same format)
- Mixed archive types (zip + tar.gz at same level)
- Depth limit enforcement (3-level zip nesting, limited to 1)
- Extension exclusion filtering
- 3-level mixed format nesting: tar.gz -> zip -> tar.gz -> files
verifying cross-format extraction and full access path chains
- Reverse format crossing: tar.gz -> zip -> tar.gz verifying
content readability at all levels and correct relationship count
Signed-off-by: Thomas Bechtold <thomasbechtold@jpberlin.de>
Integrate the archive extraction orchestrator into the SBOM creation
pipeline, enabling recursive archive cataloging when configured:
- Add Archive field to CreateSBOMConfig with WithArchiveConfig builder
- Add Exclusions field to CreateSBOMConfig with WithExclusions builder
so user --exclude patterns also filter content inside extracted
archives, not only the source filesystem
- Add setupArchiveOrchestration() that creates the orchestrator when
MaxDepth > 0, runs the iterative extract loop, and returns the
composite resolver with archive contents visible to all catalogers
- Include archive config in configuration audit trail
- Add CLI archive config section (max-depth, size/file limits,
exclude-extensions) wired through Catalog.ToArchiveConfig()
- CLI Catalog.ToSBOMConfig threads cfg.Exclusions through
WithExclusions so child resolvers respect them
When MaxDepth is 0 (default), behavior is unchanged. When > 0, archives
discovered during scanning are extracted and their contents made visible
to all catalogers via the CompositeResolver.
Signed-off-by: Thomas Bechtold <thomasbechtold@jpberlin.de>
Add tests for important orchestrator edge cases:
- Nested archives (zip-inside-zip): verifies multi-round extraction
discovers archives within previously-extracted archives
- Context cancellation: graceful handling without hanging
- Multiple archives at same level: both extracted in one round
- Corrupt archive: skipped without error, doesn't block other work
- Empty archive: not counted as extracted
- Per-archive file limit: partial extraction handled correctly
- Relationship structure: contains relationships emit the archive's
coordinates as the source side
Signed-off-by: Thomas Bechtold <thomasbechtold@jpberlin.de>
Implement the Orchestrator that manages the iterative
extract-catalog loop for recursive archive discovery:
- Scans the composite resolver for candidate archive locations,
detecting format by peeking 512 bytes via FileContentsByLocation
rather than copying every non-archive file to a probe-temp file
- Extracts identified archives to per-archive temp directories
- Builds child resolvers via an injectable ResolverFactory (kept
pluggable so tests can substitute a mock resolver without
importing syft/internal/fileresolver)
- Tracks extraction state to avoid reprocessing
- Enforces safety limits (depth, per-archive size, file count,
total extraction budget)
- Generates ContainsRelationship entries linking each archive to
its extracted-contents filesystem
- Manages temp directory lifecycle with Cleanup()
Signed-off-by: Thomas Bechtold <thomasbechtold@jpberlin.de>
Add tests for important edge cases:
- Nested children (child-in-child): verifies that archives within
archives produce correct multi-level AccessPath chains like
"/archive1.tar.gz:/inner-archive.zip:/deep-file.txt"
- Empty child resolver: no locations leak from empty archives
- Same file path in parent and child: both versions accessible
with distinct FileSystemID routing
- Context cancellation in AllLocations: no hanging
Signed-off-by: Thomas Bechtold <thomasbechtold@jpberlin.de>
Implement a file.Resolver decorator that composites a parent resolver
with child resolvers created from extracted archive contents. Files from
child resolvers appear transparently alongside parent files.
Key behaviors:
- FilesByPath/FilesByGlob/FilesByMIMEType delegate to parent AND children
- FileContentsByLocation routes to the correct resolver via FileSystemID
- Locations from children get archive-prefixed AccessPath (using ":"
separator matching Java cataloger convention) and unique FileSystemID
- Thread-safe via RWMutex for concurrent access
Signed-off-by: Thomas Bechtold <thomasbechtold@jpberlin.de>
Introduce the internal/archive package with a generic Extractor
interface for extracting archive contents to a destination directory.
Includes two implementations:
- ZipExtractor: handles zip-based archives (.zip, .jar, .war, etc.)
- TarExtractor: handles tar-based archives (.tar, .tar.gz, .tar.bz2, etc.)
Both extractors enforce configurable safety limits and protect against
common malicious-archive vectors:
- Per-archive extraction size limit
- Per-archive file count limit
- Zip-slip path traversal prevention via existing SafeJoin
- Symlink entries are written as real symlinks only when their target
resolves inside destDir; symlinks pointing outside are dropped to
avoid escape via symlink-then-write
- Non-regular non-symlink entries (devices, fifos, sockets) are skipped
because their contents can't safely round-trip to disk
Also provides FindExtractor for auto-detecting archive format, and
IsExcludedExtension for filtering by extension.
Signed-off-by: Thomas Bechtold <thomasbechtold@jpberlin.de>
Add configuration fields for generic recursive archive extraction:
- MaxDepth: maximum nesting depth for archive extraction (default 0 = disabled)
- MaxExtractionSizeBytes: per-archive extraction size limit (default 500MB)
- MaxFileCount: per-archive file count limit (default 10,000)
- MaxTotalExtractionBytes: total extraction size limit (default 2GB)
- ExcludeExtensions: archive extensions to skip during extraction
These fields lay the groundwork for cross-cataloger recursive archive
cataloging, where archives discovered during scanning can be extracted
and their contents made visible to all catalogers.
Signed-off-by: Thomas Bechtold <thomasbechtold@jpberlin.de>
Empty or whitespace-only .rockspec files cause parseRockspecBlock to
panic with "index out of range" because the existing end-of-data guard
requires len(out) > 0 before returning the "unexpected end of block"
error, letting the bare data[*i] access on the next line crash.
Split the guard so that:
- partial content at end of data still returns the existing error
- empty data (or whitespace-only) returns an empty block cleanly
Closes#4824.
Signed-off-by: Akihiko Komada <aki1770@gmail.com>
dependabot[bot]
witchcraze
Alex Goodman
Doug Clarke
anchore-oss-update-botandanchore-oss-update-bot
ChrisJr404andChris
PGrayandPGray
Rayan Salhabandcyphercodes
ChrisJr404
Ludovic Henry
Thomas Bechtold
Akihiko Komada
Sebastiaan van Stijn