18 Commits
Author SHA1 Message Date
jwinterm a43e9fe122 fix(ci): publish tag builds to the release, where they can be fetched back
Run artifacts upload fine, but this Gitea reports zero artifacts for every run
through its REST API, and the web download route answers "Not found." under
token, bearer and basic auth alike. A signed build was therefore unreachable
from anywhere except the browser session that produced it.

Release assets do work: create, upload, list and download all round-trip with
an ordinary token, so tag builds now attach their output to the tag's release
as well as to the run. Retrieval must use browser_download_url; the asset-id
API path returns the asset's JSON metadata even when asked for octet-stream.

The step is fail-safe on purpose. With no token, no resolvable release, or a
non-201 upload it explains itself and exits 0 rather than turning a green
build red. The macOS upload also gains if-no-files-found, without which an
empty publish passes silently.
2026-09-13 13:18:22 -04:00
jwinterm db151c2920 fix(ci/macos): bundle the QML modules the app actually imports
5.1.2 shipped a grey mining tab on macOS while Linux was fine. The app said
why: 'qrc:/qml/mining.qml:2:1: module "QtQuick.Layouts" is not installed',
and Contents/Resources held only appicon.icns and qt.conf, with no qml
directory at all. macdeployqt cannot scan .qrc resources, and it was called
without -qmldir, so the bundle shipped no QML modules; scene.qml survived only
because it imports nothing beyond QtQuick.

Pass -qmldir, then blanket-copy the QtQuick and QtQml trees as the AppImage
build already does, because qmlimportscanner does not follow a module's own
qmldir depends chain. Both happen before the dylib passes so new plugin deps
resolve, and before signing, which already sweeps every non-framework Mach-O.
A guard fails the build when a module is missing, because an absent QML module
is invisible until someone opens the tab.
2026-09-12 20:49:23 -04:00
jwinterm 3d38fdd135 ci(macos): bundle Qt image-plugin dylib deps (webp/sharpyuv/brotli) for clean-Mac runtime
macdeployqt bundles Qt frameworks but not the plain @rpath/*.dylib brew
deps of its image plugins, so a no-Homebrew Mac couldn't load webp. Add a
Qt-safe pass that resolves @rpath/*.dylib (never @rpath/*.framework) from
brew, copies them in, rewrites load paths, and fails the build if any
remain. Sign step now re-signs by Mach-O type (not exec bit) so the
install_name_tool-modified plugins get valid signatures.
2026-08-30 05:34:04 -04:00
jwinterm f2c3ce45a0 ci(macos): link boost statically (Boost_USE_STATIC_LIBS=ON)
The 100%-linked app dynamically referenced brew's libboost_*.dylib, which
macdeployqt failed to bundle — the app would fail to launch on any Mac
without Homebrew (notarization doesn't check runtime dylib resolution).
Static boost matches how guix/depends release builds work and removes the
runtime dependency entirely.
2026-08-30 05:08:01 -04:00
jwinterm 87d5977243 ci(macos): CPATH/LIBRARY_PATH to brew prefix so feather finds boost headers
monero's static libs built fine, but the feather target includes monero
headers pulling boost/variant.hpp, and clang doesn't search
/opt/homebrew/include by default (modern BoostConfig leaves
Boost_INCLUDE_DIRS empty).
2026-08-30 04:55:20 -04:00
jwinterm 9e6d85c246 ci(macos): disable Trezor (USE_DEVICE_TREZOR=OFF)
brew's protobuf pulls Abseil headers that require C++17, breaking the
feather compile (DEVICE_TREZOR_READY gates the protobuf link). Trezor
hardware does not support Wownero, so this removes a useless dependency
rather than working around it — the Linux lane only tolerates it because
its depends system pins an old protobuf.
2026-08-30 04:42:14 -04:00
jwinterm 0f6892c123 ci(macos): ad-hoc sign tor before embedding — arm64 macOS SIGKILLs unsigned binaries, including the runtime-unpacked copy on user machines 2026-08-29 14:17:26 -04:00
jwinterm 1fedbf0faa ci(macos): ship libevent dylib next to tor (@executable_path linkage) 2026-08-29 14:06:15 -04:00
jwinterm 410df0c8ff ci(macos): extract wownerod into an explicit dir — find(1) does not descend the /tmp symlink 2026-08-29 13:55:16 -04:00
jwinterm 1853ebf558 ci(macos): disable pipefail around yes|brew (SIGPIPE 141 when brew asks nothing) 2026-08-29 13:47:37 -04:00
jwinterm a2b34f5ebc ci(macos): embed tor (expert bundle 15.0) — arms daemon --tx-proxy via HAS_TOR_BIN, parity with linux lane 2026-08-29 13:44:28 -04:00
jwinterm 8b846a8dc0 ci(macos): pin CMAKE_PREFIX_PATH to brew qt6 (qt@5 keg coexists unlinked on the runner) 2026-08-29 13:37:41 -04:00
jwinterm dc5f940801 ci(macos): pipe yes into brew — Homebrew 6 prompts for dependency installs regardless of NONINTERACTIVE 2026-08-29 13:27:05 -04:00
jwinterm 36d878e7bf ci(macos): also skip dependent upgrades in brew install 2026-08-29 13:16:13 -04:00
jwinterm a526061d97 ci(macos): place appicon.icns into the bundle 2026-08-29 13:15:24 -04:00
jwinterm f4c5ec3c25 ci(macos): non-interactive brew 2026-08-29 13:10:03 -04:00
jwinterm 1217ce174c ci(macos): sign + notarize with Developer ID cert via desktop-lane secrets 2026-08-29 12:17:21 -04:00
jwinterm 914fa1ed32 ci: native macOS lane for the Gitea self-hosted runner
Dispatch-only workflow in .gitea/workflows (GitHub ignores the
directory): brew dependencies, native cmake build with the embedded
wownerod fetched per-arch, unsigned app bundle artifact. Runs on the
mac runner whose work dir lives on DataSSD and is outside every backup
path. Signing/notarization follows once a Developer ID Application
certificate is provisioned as repo secrets.
2026-08-29 07:04:39 -04:00