trojita-git

• Path: arch/maintained/trojita-git
• User (Andreas Baumann) is the upstream AUR maintainer for this pkgbase AND for the separate trojita-qt5-git pkgbase (see memory/trojita-qt5-git.md) – two genuinely different packages: this one is Qt6 (WITH_WEBKIT=OFF), trojita-qt5-git is Qt5+WebKit.

2026-08-20: fixed a self-inflicted mixup, this is the Qt6 variant

• The mess: earlier this session (2026-08-17), this directory’s PKGBUILD got overwritten to match trojita-qt5-git’s content instead of its own real upstream. The giveaway that went unnoticed at the time: both AUR pkgbases (trojita-git and trojita-qt5-git) literally set pkgname=trojita-qt5-git inside their own PKGBUILD – an upstream copy-paste inconsistency on the user’s own AUR side, which made it easy to mistake one for the other. When a cascade of missing Qt6 dependencies showed up during an earlier build attempt here, that was misread as “this qt5-named package is wrongly building against Qt6” and “fixed” by overwriting this directory with trojita-qt5-git’s actual (Qt5+WebKit) content – backwards: this pkgbase is genuinely supposed to be Qt6, the cascade was just missing real dependencies, not evidence of drift.
• Fix: restored from real upstream (cgit/aur.git/plain/PKGBUILD?h=trojita-git) verbatim, with one local deviation: pkgname corrected from trojita-qt5-git (the upstream copy-paste bug) to trojita-git, matching this directory’s actual identity. pkgdesc fits either way (“A fast QT IMAP e-mail client”, no Qt5/webkit-specific claim – verified consistent with the Qt6/WITH_WEBKIT=OFF build).
• Real missing dependencies, found and fixed one at a time by actually building (not chasing forever – each was a genuine gap in upstream’s own depends=()):
  1. qt6-5compat (provides Qt6Core5Compat, -- Could NOT find Qt6Core5Compat).
  2. kitemmodels (KF6 flavor, provides KF6ItemModels).
  3. qtkeychain -> qtkeychain-qt6 (upstream’s depends is stale on the pre-Qt6-split name, same class of fix as qtkeychain-qt5 in trojita-qt5-git).
  4. -DCMAKE_PREFIX_PATH=/usr/lib/cmake/QGpgmeQt5 -> QGpgmeQt6 (upstream’s own build() still pointed at the Qt5 gpgme cmake config even though this is the Qt6 build).
• i686 dropped: kitemmodels (KF6 flavor) isn’t ported to i686 on archlinux32 – only the old KF5-era kitemmodels5-lineage package exists there (confirmed via the official archlinux32 mirror listing, extra/kitemmodels-5.105.0-1.0-i686.pkg.tar.zst, no KF6 version). Unsatisfiable, not fixable locally. arch=(x86_64) only. User confirmed not to chase this further – KF6/Qt6 generally isn’t working on archlinux32 yet.
• pkgrel history: 1 (real upstream content, restored) -> 2 (qt6-5compat) -> 3 (kitemmodels) -> 4 (qtkeychain-qt6 + QGpgmeQt6 path + i686 drop) -> auto-reset to 1 by the -git pkgver recompute picking up a newer upstream commit during the final successful build (0.7.r1004.g80ee0085 vs the r839 pinned earlier) – expected -git behavior, not a regression.
• End state: x86_64 builds and publishes cleanly.