mdcat

Build times (auto-updated)

• x86_64: 19m39s (2026-10-01 16:46 UTC, MAKEFLAGS=-j6)

---

Not yet vendored under arch/ (as of 2026-09-19). Tracking notes ahead of adoption so nothing gets lost.

Background

• Upstream swsnr/mdcat was removed from official Arch repos (unmaintained upstream). The AUR mdcat package now tracks a fork, BIRSAx2/mdcat, which is actively maintained.
• User (andreas_baumann) was AUR submitter and, per their own AUR comment on 2026-02-17, merged/fixed and republished it at that time (“I just merged the AUR and package git and published it back. Currently it builds with minimal patching.”).
• As of 2026-09-19 the AUR page shows Maintainer: None (orphaned) – RPC confirms "Maintainer": null. scripts/add_maintained.sh mdcat correctly refuses to vendor it under maintained/ for this reason (we’re not listed as maintainer or co-maintainer).
• 2026-09-19: user filed an AUR orphan/adoption request for mdcat. Pending Trusted User review – may take a while. Do not run add_maintained.sh mdcat until the adoption is actually approved (AUR page shows a real maintainer name again).

Known issue – FIXED locally 2026-09-19, not yet pushed

• 2026-09-16, reported by AUR user ZEBofx: check() runs cargo test, which exercises the mdpick fuzzy-picker path and spawns less as a pager – but less isn’t listed in checkdepends, so makepkg fails in a clean chroot (no less preinstalled) with: Error: Failed to spawn pager less with args ["-r"]: No such file or directory (os error 2).
• User confirmed the failure themselves on the AUR thread 2026-09-16.
• Note: optdepends already listed less: for mdless, but that’s purely advisory for end users – makepkg/makechrootpkg never installs optdepends into the build chroot, only depends/makedepends/checkdepends. That’s exactly why the clean chroot lacked less despite it being “declared”.
• Checked the test source (tests/cli.rs): fzf is stubbed with a fake sh script in the test sandbox, so it does NOT need to be a real checkdepend – only less is spawned for real.
• 2026-09-19: re-cloned fresh into /data/INSTALL/mdcat (the old checkout there was stale, HEAD at 2.16.0 while AUR’s actual master was already at 2.17.0-1 – re-cloning avoided accidentally reverting upstream on push). Added checkdepends=('less'), regenerated .SRCINFO, updated the # Maintainer: header to Andreas Baumann (former maintainers moved to # Contributor: lines, standard AUR takeover convention). Test-built clean via repo_build_staged.sh (x86_64, REPO_BUILD_ONLY_ARCH=x86_64): all 39 tests pass, including mdpick::renders_the_file_selected_in_fzf ... ok – confirms the fix actually resolves the reported failure.
• Still NOT pushed to AUR and NOT vendored under arch/maintained/ – waiting on the pending adoption request (AUR page still shows Maintainer: None as of this check). Once adoption is approved: git -C /data/INSTALL/mdcat push (or move it into arch/maintained/mdcat via add_maintained.sh first, then push from there – either order works, just don’t lose the local checkdepends fix in the process), then the normal maintained/ flow going forward.

2026-09-27: adoption completed, published for the first time

AUR adoption went through (RPC now shows Maintainer: andreas_baumann) and a real submodule now exists at arch/maintained/mdcat. Found it staged with the Maintainer: header and pkgrel (1 -> 2) already updated, but missing the actual checkdepends=('less') fix documented above – likely lost when re-cloning fresh partway through that earlier investigation. Re-added it before building; without it this would have failed check() again with the exact same ZEBofx-reported error.

i686 fails for an unrelated, unfixable-here reason: cargo fetch --locked pulls idna_adapter 1.2.2 (a transitive dependency), whose Cargo.toml requires the edition2024 Cargo feature – not stabilized in archlinux32’s own official i686 rust package (1:1.81.0-1.0, built September 2024, confirmed via pacman -Si and Packager: TasosSah – a genuine upstream toolchain-lag issue in the third-party archlinux32 project itself, not something in our own PKGBUILD or infrastructure). x86_64’s rust/cargo is current and unaffected. Not fixed – would need archlinux32 itself to update its rust package, out of scope. arch=('i686' 'x86_64') left as-is (matches upstream); i686 just won’t build until that catches up.

Built + published clean, x86_64 only, 2.17.0-2. All 39 tests passed including the mdpick/less-spawning one the fix targets.

2026-09-28: routine check – fixed a stale .nvchecker.toml tracking the wrong (dead) upstream

Asked to “check and update mdcat”. Upstream’s real tags (git ls-remote --tags against BIRSAx2/mdcat) top out at 2.17.0 – matches the published pkgver exactly, so no version bump needed.

scripts/check_for_updates_maintained.sh --only maintained/mdcat reported OK ... 2.17.0 2.7.1, which looked suspicious (that second number should be the detected upstream version, and 2.7.1 is far below our 2.17.0) – root-caused to .nvchecker.toml still pointing git = "https://github.com/swsnr/mdcat.git", the original abandoned upstream (see the “Background” section above – this project switched to tracking the BIRSAx2/mdcat fork a while ago, but apparently only the PKGBUILD’s own source= URL got updated at the time, not the nvchecker config). 2.7.1 is just the last tag that dead repo ever got. The check only reported “OK” by coincidence (our real version numerically exceeds the stale target it was comparing against) – it was never actually validating against real current upstream, and would silently miss a real new BIRSAx2 release forever (vercmp would never see it as newer than a frozen 2.7.1). Fixed by pointing .nvchecker.toml at BIRSAx2/mdcat – re-ran the check, now correctly reports 2.17.0 2.17.0. Also added an STATES.txt entry for this package (never had one – vendored after the last full check run).

2026-10-01: routine bump 2.17.0 -> 2.18.0

Full update_cycle.sh UPD flag. Upstream tarball + sha256 fetched clean via updpkgsums. Dependency tree has grown substantially since 2.17.0 – now pulls in resvg/usvg/tiny-skia/ratatui-image etc. (image/SVG rendering in the terminal, presumably inline-image support), ~300+ crates vs. a much smaller set before. Built x86_64-only as usual (i686 still blocked by archlinux32’s stale rust package, see below). All 39 tests passed, built+signed+published cleanly, old 2.17.0-2 archived.

Current AUR PKGBUILD snapshot (2.17.0-1, for reference)

• depends=(gcc-libs openssl curl), makedepends=(asciidoctor cargo), optdepends=(less fzf) – no checkdepends array at all currently.
• Builds via cargo build --frozen --release, tests via cargo test --frozen, man page via asciidoctor -b manpage.
• Locally installed version on this machine is older (2.10.1-1, built 2026-07-07) – once adopted, rebuild to current (2.17.0 as of this writing) via the normal maintained/ flow.