rxvt

Build times (auto-updated)

• x86_64: 25s (2026-09-11 10:05 UTC, MAKEFLAGS=-j6)
• pentium4: 25s (2026-09-11 10:05 UTC, MAKEFLAGS=-j6)
• i686: 26s (2026-09-11 10:05 UTC, MAKEFLAGS=-j6)
• i486: 21s (2026-09-11 10:06 UTC, MAKEFLAGS=-j6)
• armv6h: 6m38s (2026-09-11 10:13 UTC, remote, serial)
• armv7h: 2m54s (2026-09-11 10:13 UTC, remote, MAKEFLAGS=-j4)
• aarch64: 58s (2026-09-11 13:45 UTC, remote, MAKEFLAGS=-j4)

---

• Path: arch/maintained/rxvt

2026-09-11: expanded arch=() to all 7 archs, fixed aarch64’s ancient config.guess/config.sub

Was arch=('i686' 'x86_64') only. Expanded to the full x86_64 pentium4 i686 i486 armv6h armv7h aarch64 set – purely additive, no pkgrel bump for that part (see memory/PKGREL_VS_PKGVER.md). x86_64/pentium4/i686/i486/armv6h/armv7h all built and published cleanly first try.

aarch64 failed: configure: error: cannot guess build type; you must specify one. Root cause: the bundled autoconf/config.guess (and config.sub) are dated 2002-09-03 – predate the aarch64 architecture entirely (grepped both files, zero mentions of “aarch64”), so configure’s platform auto-detection can’t identify the board, and even an explicit --build=aarch64-... override would still fail since config.sub itself doesn’t recognize the string either.

Fix: pkgrel 10 -> 11, prepare() now does (arch-conditional, x86_64/etc. untouched since their 2002-era config.guess already handles them fine):

if [ "$CARCH" = "aarch64" ]; then
  cp /usr/share/automake-*/config.{guess,sub} autoconf/
fi

automake is part of base-devel, already present in every chroot and confirmed present on eurobuild14 too – no new makedepends needed.

First attempt at this fix failed too (same “2002-09-03” error, confirmed by ssh’ing onto eurobuild14 and checking the actual extracted source’s autoconf/config.guess timestamp post-failure) – copied the fresh files into . instead of autoconf/, the actual directory configure’s ac_aux_dir search resolves to for this package (confirmed via grep ac_config_guess configure). Fixed by targeting the right destination; retried and aarch64 published cleanly.

User’s suggestion mid-fix was a full autoreconf -fi instead – not used, since the narrower config.guess/config.sub swap already targets the exact failure and a full regeneration risks disturbing the several sed edits build() already applies against this specific old configure-generated config.h layout.