| Age | Commit message (Collapse) | Author | Files | Lines |
|
Slackware -current ships kddockwidgets 2.4.1 in the KDE series
(slackware64/kde), the same version the SBo tree carries, so listing it
as a dependency made the build pull in and rebuild a library the base
system already provides. This package targets -current only, so the
dependency is always redundant here.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
|
Port the 2.4.0 script from Slackware 15.0 to -current. 15.0 shipped none
of the libraries pcsx2 needs, so the old script vendored sixteen sources
and patched upstream's dependency builder with hardcoded sed line ranges
to skip the Qt build. -current already provides SDL3, Qt6, libpng,
libjpeg-turbo, libwebp, lz4, zstd, ffmpeg and the Vulkan headers at or
above the versions upstream pins, and kddockwidgets 2.4.1 comes from SBo,
so both the vendoring and the line-range surgery mostly go away.
Five dependencies have no packaged counterpart and are still built here,
now directly with upstream's own flags rather than through the patched
builder: libbacktrace, plutovg, plutosvg, rapidyaml and shaderc (with its
three pinned third-party trees). They are built static into a private
prefix that only the build sees, so nothing lands in the package and the
old /usr/lib64/pcsx2 libdir and its LD_LIBRARY_PATH desktop wrapper are
both gone. Verified: the package ships one binary and no shared objects,
no NEEDED entry references the vendored five, and every remaining NEEDED
resolves in a clean -current container with only kddockwidgets added.
pcsx2 2.8.2 still reads AVCodec::pix_fmts and sample_fmts, removed in
ffmpeg 8, so building against -current's ffmpeg 9 fails in GSCapture.cpp.
Disable find_package(FFMPEG) to select the bundled headers upstream ships
for this case; ffmpeg is dlopen'd at runtime, so capture still uses the
system library.
Also drop the llvm-opt dependency now that -current's llvm provides
clang, restrict ARCH to x86_64 with an explicit error, and remove
qt-6.10.patch, which patched the no longer vendored KDDockWidgets 2.2.3.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
src/dbus/CMakeLists.txt builds up DBUS_INTERFACES with its generated
qt_add_dbus_interface sources, then calls add_subdirectory(dbusmenu) at the
bottom. The subdirectory inherits the variable and qt_add_dbus_interface
appends to it, so quickshell-dbusmenu ends up listing src/dbus's generated
sources as well. Those entries are relative filenames, so make resolves them
against the dbusmenu binary dir and the build dies with:
No rule to make target 'src/dbus/dbus_objectmanager.cpp', needed by
'src/dbus/dbusmenu/CMakeFiles/quickshell-dbusmenu_autogen_timestamp_deps'
Unset DBUS_INTERFACES before the add_subdirectory so dbusmenu starts clean.
Still unfixed upstream as of master.
Verified with a containerized test build: SUCCESS, lint clean.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XLAYexNKJ4BKdyjiLiqdMz
|
|
Add the quickshell package, a QtQuick toolkit for building status bars,
widgets and other desktop components.
Bumped from the initially added 0.2.1 to 0.3.1. Upstream renamed the
cmake option CRASH_REPORTER to CRASH_HANDLER in 0.3.x, so the old flag
would have been silently ignored and pulled in the crash handler along
with its cpptrace dependency.
Other fixes over the initial script:
- DOWNLOAD was UNSUPPORTED with the real URL in DOWNLOAD_x86_64, which
wrongly marked every non-x86_64 arch as unbuildable. The source is
arch independent, so it belongs in DOWNLOAD.
- REQUIRES said "qt6 cli11". qt6 is not an SBo package, and the cli11
entry did not match the SBo package name, which is CLI11.
- The gitea archive is served under its Content-Disposition name
(quickshell-v0.3.1.tar.gz) rather than the URL basename, so accept
either filename when extracting.
- DISTRIBUTOR claimed SlackBuilds.org. Upstream asks that this string
identify the specific package source, and it is compiled into the
binary, so name this repo instead.
- Align TAG and PKGTYPE with the rest of the repo (_SBo, tgz), fix the
slack-desc handy ruler to 72 characters, add the missing README and
drop the leftover template comments.
Builds clean on -current with a clean sbopkglint. Slackware 15.0 ships
Qt5 only, so the build fails there on a missing Qt6; the package is
-current only and the README notes this.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01456ddh6BkCimbiPi5WucyT
|
|
llama.cpp-vulkan now uses versioned release tags, so sbolint no longer
misreads the tag as a commit hash and the SBOLINT=no exception is gone.
The nvchecker --local-deps finding was a symptom of a general rule, in-repo
sibling deps never resolve from the SBo tree, so piper-tts hit it too.
Reworded to cover any such package.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
QTMAILDIR_BUILD_NUMBER is ON upstream so developers can distinguish
builds of an unreleased version. In a package it makes the release
report "0.27.0 build 1", numbering a release after a counter that
restarts in every fresh build directory. Turn it off.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yeUGcX28FHJNQMtkzqx5e
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|