Version: git-pkgs/registries current main at 418e68a.
Trying to create an Arch Linux repository client after importing all registries currently returns unknown ecosystem: alpm:
_, err := registries.New("alpm", "https://geo.mirror.pkgbuild.com/core/os/x86_64/core.db", nil)Add an alpm registry that reads a configured pacman sync database. The URL should identify one concrete repository database, including repository and architecture, such as Arch's current core.db. There is no default ALPM repository. Register an empty default and make core.New return a clear error when the URL is missing.
A sync database is a compressed tar archive with one directory per package version and desc, depends, or files entries. Pacman's reader accepts the fields in those entries and maps NAME, VERSION, FILENAME, DESC, URL, LICENSE, ARCH, BUILDDATE, PACKAGER, SHA256SUM, and dependency groups. repo-add can generate small test databases from package fixtures.
Map package description, homepage, licenses, architecture, build date, packager, compressed and installed sizes, groups, base package, filename, and SHA-256 checksum. Resolve the package download URL from FILENAME relative to the database directory. Preserve the full version, including epoch and package release, without SemVer normalization.
FetchDependencies should parse DEPENDS requirements as runtime dependencies and OPTDEPENDS as optional dependencies, removing only the optional dependency's human-readable description. Keep make, check, conflict, replace, and provide relationships in metadata until the shared dependency type can name those roles. Parse PACKAGER values such as Name <email> into maintainer fields while retaining unstructured values.
The alpm PURL type requires a vendor namespace and supports an architecture qualifier. When called through NewFromPURL, split a full name such as arch/pacman for database lookup, retain arch as Package.Namespace, and keep architecture in version metadata. The configured database URL remains the source of repository and architecture selection because the current registry interface does not pass qualifiers to FetchPackage.
Handle the gzip-compressed tar format used by the current core.db example and the xz or zstd tar variants pacman can read, without shelling out to pacman or bsdtar. Reject unsafe archive paths and malformed required records.
Follow the new ecosystem layout, register the package from all/all.go, and add it to the root ecosystem tests.
Tests should cover compressed databases, epoch and package-release versions, any and concrete architectures, vendor namespaces, required and optional dependencies, packager parsing, SHA-256 integrity, resolved package URLs, missing optional fields, malformed archives and entries, a missing configured URL, authentication through the supplied HTTP client, and not-found packages and versions.
Version:
git-pkgs/registriescurrentmainat418e68a.Trying to create an Arch Linux repository client after importing all registries currently returns
unknown ecosystem: alpm:Add an
alpmregistry that reads a configured pacman sync database. The URL should identify one concrete repository database, including repository and architecture, such as Arch's currentcore.db. There is no default ALPM repository. Register an empty default and makecore.Newreturn a clear error when the URL is missing.A sync database is a compressed tar archive with one directory per package version and
desc,depends, orfilesentries. Pacman's reader accepts the fields in those entries and mapsNAME,VERSION,FILENAME,DESC,URL,LICENSE,ARCH,BUILDDATE,PACKAGER,SHA256SUM, and dependency groups.repo-addcan generate small test databases from package fixtures.Map package description, homepage, licenses, architecture, build date, packager, compressed and installed sizes, groups, base package, filename, and SHA-256 checksum. Resolve the package download URL from
FILENAMErelative to the database directory. Preserve the full version, including epoch and package release, without SemVer normalization.FetchDependenciesshould parseDEPENDSrequirements as runtime dependencies andOPTDEPENDSas optional dependencies, removing only the optional dependency's human-readable description. Keep make, check, conflict, replace, and provide relationships in metadata until the shared dependency type can name those roles. ParsePACKAGERvalues such asName <email>into maintainer fields while retaining unstructured values.The
alpmPURL type requires a vendor namespace and supports an architecture qualifier. When called throughNewFromPURL, split a full name such asarch/pacmanfor database lookup, retainarchasPackage.Namespace, and keep architecture in version metadata. The configured database URL remains the source of repository and architecture selection because the current registry interface does not pass qualifiers toFetchPackage.Handle the gzip-compressed tar format used by the current
core.dbexample and the xz or zstd tar variants pacman can read, without shelling out to pacman orbsdtar. Reject unsafe archive paths and malformed required records.Follow the new ecosystem layout, register the package from
all/all.go, and add it to the root ecosystem tests.Tests should cover compressed databases, epoch and package-release versions,
anyand concrete architectures, vendor namespaces, required and optional dependencies, packager parsing, SHA-256 integrity, resolved package URLs, missing optional fields, malformed archives and entries, a missing configured URL, authentication through the supplied HTTP client, and not-found packages and versions.