archlinux-s390x/scripts
Mathieu Benoit e137999f8b [FIX] the xar cycle reformed inside stage 2
39 of 39 packages failed on the same line:

  bsdtar: error while loading shared libraries: libicuuc.so.76

libarchive is in STAGE2_SKIP_HOOKS, so stage 2 rebuilt it as Arch does -- with
xar, linking whatever libxml2 was installed at that moment: the stage-1 one,
which wants the host's ICU 76 while ours ships 78. bsdtar is what makepkg uses
to extract sources and write packages, so nothing could come out, including
libxml2 itself. The stage-1 fix does not reach in here.

No new mechanism was needed. libarchive is dropped from stage2.state, so the
restore falls back to the stage-1 build that has no xar and therefore no libxml2
at all; libxml2 is hoisted and rebuilt against our icu; libarchive is rebuilt
later with xar, against a libxml2 whose ICU now agrees. Verified: bsdtar 3.8.9
runs, and libxml2 is the first OK of the pass.

The libgpgme.so.11 errors in five logs were noise -- makepkg carries on past a
failed .BUILDINFO. Reading the FIRST error in a log had attributed five failures
to the wrong cause; the last error before the abort is the one that matters.

--- FR ---

39 paquets sur 39 ont échoué sur la même ligne :

  bsdtar: error while loading shared libraries: libicuuc.so.76

libarchive figure dans STAGE2_SKIP_HOOKS : l'étage 2 l'a donc rebâti comme le
fait Arch — avec xar, contre le libxml2 installé à cet instant, celui de
l'étage 1, qui veut l'ICU 76 de l'hôte quand le nôtre livre la 78. bsdtar est ce
dont makepkg se sert pour extraire et pour écrire les paquets : plus rien ne
sortait, libxml2 compris. Le correctif d'étage 1 n'atteint pas l'intérieur.

Aucun mécanisme nouveau. libarchive quitte stage2.state, donc la restauration
retombe sur la construction d'étage 1, sans xar et donc sans libxml2 ; libxml2
est hissé et rebâti contre notre icu ; libarchive est rebâti plus tard avec xar,
contre un libxml2 dont l'ICU concorde. Vérifié : bsdtar 3.8.9 démarre, et
libxml2 est le premier OK de la passe.

Les erreurs libgpgme.so.11 de cinq journaux étaient du bruit — makepkg poursuit
après un .BUILDINFO manqué. Lire la PREMIÈRE erreur d'un journal avait attribué
cinq échecs à la mauvaise cause ; c'est la dernière avant l'abandon qui compte.

Assisted-by: Claude Opus 5
2026-08-21 02:18:25 -04:00
..
devtools [FIX] libdir: the Debian host hid the libraries from Arch 2026-08-17 01:33:41 -04:00
bootstrap-pacman.sh [FIX] stage 2: the build tools the host was providing silently 2026-08-20 01:34:41 -04:00
build-arch-kernel.sh making the house a home 2026-01-27 23:26:31 +01:00
build-busybox-zvm.sh making the house a home 2026-01-27 23:26:31 +01:00
build-container.sh making the house a home 2026-01-27 23:26:31 +01:00
build-initramfs-final.sh making the house a home 2026-01-27 23:26:31 +01:00
build-stage1.sh [FIX] stage 2: the build tools the host was providing silently 2026-08-20 01:34:41 -04:00
build-stage2.sh [FIX] the xar cycle reformed inside stage 2 2026-08-21 02:18:25 -04:00
build-systemd-zvm.sh making the house a home 2026-01-27 23:26:31 +01:00
common.sh cleaned up scripts to be more DRY 2025-06-05 20:48:16 +02:00
create-rootfs.sh making the house a home 2026-01-27 23:26:31 +01:00
deploy-and-build-systemd-zvm.sh cleaned up scripts to be more DRY 2025-06-05 20:48:16 +02:00
lib-watch.sh [FIX] stage 2: the build tools the host was providing silently 2026-08-20 01:34:41 -04:00
packages.sh [FIX] help2man was uninstallable, which stopped the whole pass 2026-08-21 00:33:07 -04:00
prepare-boot.sh making the house a home 2026-01-27 23:26:31 +01:00
test-chroot.sh [FIX] one stray .a file broke every pkg-config lookup 2026-08-21 00:47:09 -04:00
test-qemu.sh cleaned up scripts to be more DRY 2025-06-05 20:48:16 +02:00
zvm-utils.sh cleaned up scripts to be more DRY 2025-06-05 20:48:16 +02:00