basic256

Build times (auto-updated)

• i686: 2m49s (2026-09-17 18:26 UTC, MAKEFLAGS=-j6)
• x86_64: 1m25s (2026-09-18 08:27 UTC, MAKEFLAGS=-j6)
• pentium4: 3m23s (2026-09-17 18:32 UTC, MAKEFLAGS=-j6)

---

• Path: arch/maintained/basic256
• PKGBUILD already listed arch=('i686' 'x86_64'), but only x86_64 had ever actually been built/published to the private repo (confirmed via scripts/repo_arch_status.sh: i686 showed “not published”).

2026-09-17: extended to i686 – icu76 linking gap (same pattern as qgpgme-qt5/trojita-qt5-git)

• First i686 build attempt failed linking basic256 itself: undefined reference to 'ucnv_...76' etc. Root cause: the i686 chroot’s system Qt5Core.so links against the legacy icu76 soname, but nothing in the dependency chain declares it, so the newer default icu (78.3) is what’s actually installed. Identical root cause already seen and fixed for qgpgme-qt5 / trojita-qt5-git (see those memory files) – icu76 itself is available prebuilt in the archlinuxaba repo for i686.
• Fix: added makedepends_i686=('icu76'), bumped pkgrel 3 -> 4 (packaging-only fix), regenerated .SRCINFO (which was stale anyway – still listed pkgrel=2 and the old espeak depend, both already gone from PKGBUILD).
• Rebuild: i686 built and signed successfully with the fix.
• x86_64 rebuild (needed too, since pkgrel bumped) FAILED in the same run, but only because archlinuxaba.db sync timed out – the private repo host (euroweb.lan / archlinux32.andreasbaumann.cc) was unreachable (“No route to host”) at that moment. Unrelated to the icu76 fix; not yet retried.
• Publish also failed for the same reason: repo_publish.sh couldn’t reach root@euroweb.lan to create os/i686 under the mirror.
• Host (euroweb.lan / archlinux32.andreasbaumann.cc) came back up shortly after; retried and both i686 and x86_64 built and published cleanly at 2.0.99.10-4 (basic256 + basic256-debug, both archs). Confirmed via scripts/repo_arch_status.sh: both archs show “published (2.0.99.10-4)”.

2026-09-17: further extended to i486/pentium4 – i486 blocked on missing icu72

• Added i486/pentium4 to arch=(), plus makedepends_i486=('icu76') / makedepends_pentium4=('icu76') up front (same Qt5/icu chain issue is arch-independent, so pre-empted it rather than waiting for a guaranteed failure). pkgrel 4 -> 5.
• pentium4 built and published cleanly with icu76, same as i686.
• i486 FAILED differently: undefined reference to 'ucnv_..._72' etc – i486’s system Qt5Core.so was built against icu72 specifically, not icu76. Checked every configured repo section (archlinuxaba/core/extra) for i486: no icu72 package exists at all (only icu74/75/76 are available). This is the same arch-vintage-drift issue as memory/trojita-qt5-git.md’s icu72/ libxml2-legacy findings – different archlinux32 arches’ Qt5 stacks were rebuilt against different legacy ICU versions at different times, and not every legacy version got carried forward for every arch.
• Building icu72 from source for i486 ourselves (like gpgme/ gpgmepp were built for the trojita chain) would be a real side-project, not a quick fix – user’s call: leave i486 in arch=() as-is (documented broken here) rather than chase it now. i486 stays unpublished; i686/x86_64/pentium4 all publish cleanly at 2.0.99.10-5.
• If picked up later: start by trying to build/package icu72 for i486 the same way gpgme/gpgmepp were done for trojita (see memory/gpgme.md/memory/gpgmepp.md for that pattern), then retry this package’s i486 build.

2026-09-17: i486 dropped for good – user decision

• User: don’t chase building against such an old Qt/icu72 combination for i486. Removed i486 from arch=() entirely (and its now-dead makedepends_i486=('icu76')), rather than leaving it declared-but- broken. pkgrel 5 -> 6.
• Rebuilt/republished i686/x86_64/pentium4 at 2.0.99.10-6 (a pure arch=()/metadata change, no functional difference to those three, but pkgrel bumped anyway to keep PKGBUILD/.SRCINFO/published state consistent – same convention followed when trojita-qt5-git similarly reverted a 32-bit arch=() addition). All three confirmed published (2.0.99.10-6) via scripts/repo_arch_status.sh.
• Final supported archs for this package: i686, x86_64, pentium4. Not i486 – don’t re-attempt without first solving the icu72 availability gap noted above.

2026-09-18: upstream v2.2.0 – moved to GitHub, CMake, Qt6-only; split into basic256 (Qt6) + basic256-qt5 (last Qt5 release)

• AUR flagged out-of-date; comment from dkorzhevin: “BASIC-256 v2.2.0 released”. Verified against real upstream, not just the comment (see feedback_verify_aur_flags_before_acting.md):

  • Our .nvchecker.toml tracked SourceForge’s kidbasic project, which is frozen at 2.0.99.10 – that project is abandoned.
  • Real upstream is now github.com/uglymike17/basic256 (basic256.org links there), tagged releases up to v2.2.0.
  • Build system moved from qmake (BASIC256.pro) to CMake.
  • Qt6 is mandatory as of the earliest GitHub tag (v2.1.beta.01) – CMakeLists.txt does find_package(QT NAMES Qt6 REQUIRED ...), and upstream’s own COMPILING.txt says “Qt5 is no longer supported.” Confirmed by checking every tag’s CMakeLists.txt, not just main.
  • License changed GPL2 -> GPL3 (GPL-3.0-or-later, confirmed via license.txt + the “or (at your option) any later version” wording in Version.h).
  • Needed Qt6 components: Core, Gui, Widgets, Sql, PrintSupport, SerialPort, Multimedia, TextToSpeech, LinguistTools, Network, plus qt6-declarative as a build-time-only dep (Qt6TextToSpeech’s own CMake package config requires Qt6QmlIntegration even though this project has no QML of its own – not a runtime dependency, pacman -Qi qt6-speech doesn’t list it).
  • Checked archlinux32’s extra repo for i686/pentium4: all the above Qt6 components are available there too, so 32-bit isn’t a dead end for the Qt6 line if picked up later.

• basic256 (this dir) rewritten for v2.2.0: new GitHub tarball source, cmake/ninja-free plain cmake -B build + cmake --build build(), Qt6 deps, GPL-3.0-or-later. package() uses DESTDIR=... cmake --install build for the binary, then manually installs Modules/*.kbs next to the binary (/usr/bin/Modules/, not /usr/share/...) – upstream’s own interpreter resolves include’d modules via QCoreApplication::applicationDirPath(), not a fixed system data dir (see LEX/basicParse.l), so this is upstream’s actual runtime expectation, not a packaging choice; not worth patching around for now. Examples/icon/desktop installed manually as before (upstream’s own install() only handles the binary). Built and published for x86_64 only so far (2.2.0-1), per explicit user decision – i686/pentium4 not yet attempted for the Qt6 line; re-check archlinux32’s Qt6 packages there before trying (should be fine per the check above, but unverified by an actual build).

• Old qmake+Qt5 PKGBUILD (last state: 2.0.99.10-6, all three archs) is preserved as-is in git history (commit 4fa4564) – reused verbatim, just renamed, for the new package below.

• New package basic256-qt5 (arch/maintained/basic256-qt5): user wanted a fallback for systems without Qt6, using “the last version which still builds with Qt5” – turned out to be exactly our own already-working old package (SourceForge 2.0.99.10, qmake, Qt5), not some newer untagged commit from the GitHub repo (checked: even GitHub’s earliest tag is already Qt6-only, so there’s no newer Qt5 release to bisect down to – SourceForge’s 2.0.99.10 IS the last Qt5 release). pkgname=basic256-qt5, pkgrel reset to 1 (new package identity), provides=('basic256') / conflicts=('basic256') since it installs the same /usr/bin/basic256 – same pattern as trojita-qt5-git’s provides=('trojita')/conflicts=(...). Built and published cleanly on i686/x86_64/pentium4 at 2.0.99.10-1 (identical content to old basic256’s working icu76-fixed build, just renamed) – no new build issues since nothing about the actual package changed.

• This is a brand-new package, not yet an AUR submodule/git repo of its own – created as a plain directory under arch/maintained/. Needs the user’s own git init/AUR remote add/push to actually submit it to AUR (never done by me – see feedback_never_git_commit_or_push.md); until then it only exists in our private repo.