The segfault: Qt destroys and recreates a widget's QOpenGLContext whenever it
moves to another top-level window, then calls initializeGL() again. libmpv
permits one render context per handle, so the second creation failed with
"There is already a mpv_render_context set" -- and the except branch assigned
self._render_ctx = None, dropping the last Python reference to the first
context, which libmpv was still holding a function pointer into. python-mpv's
MpvRenderContext has no __del__ and free() does not unregister the callback,
so the ctypes trampoline was collected while registered and the next frame
notification jumped into freed memory.
Two invariants fix it. initializeGL() now tears down any existing render
context first, so a second call is an ordinary recreation. Teardown clears
update_cb, calls free() with the GL context current, and only then drops the
reference -- and it is connected to QOpenGLContext.aboutToBeDestroyed, so it
runs before the GL context dies instead of never. The local reference during
teardown is load-bearing: it is what keeps the trampoline alive until free()
returns.
Three things were destroying that context. Fullscreen reparented the panel
into a new top-level window (twice per toggle) and put it back at the end of
the splitter, losing the pane layout; it now fullscreens the main window and
hides the chrome, reparenting nothing. The tab cross-fade and the dialog blur
both grab() the widget tree, which on an OpenGL surface forces a framebuffer
readback and returns black -- the fade is skipped for pages holding the video,
and dialogs dim rather than blur.
Verified on a real Wayland GL context: ten forced context destroy/recreate
cycles re-establish the render context every time, and the full app survives
tab switching, six fullscreen toggles and resizes with no render-context
error and a clean exit. Before this, the same startup dumped core.
Also here because they are one-line consequences of touching _create_mpv: an
explicit per-platform hwdec list ending in software decoding, and the restored
volume actually reaching mpv -- the slider set its value before connecting its
signal, so playback always started at 100.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The update check asked PyPI for the `ytsage` package's version, asked GitHub
for oop7/YTSage's release notes, and pointed the download button at upstream's
releases. None of that describes this program. Being reminded to install
YTSage was the visible symptom; the cause was that the check had never been
repointed when the fork was made.
It now reads SageTube's own releases from git.houmeres.sk. Gitea's release API
is shaped like GitHub's, so the dialog and the caller are unchanged -- the
class keeps its name and signal signature, and only its body moved out to
core/ytsage_app_update.py, which is fork-owned and will not conflict on the
next merge from upstream.
Upstream's inherited tags end in `b` (v5.3.0b and earlier). packaging reads
that as a beta marker, so they sort below v5.4.0 and a stable instance cannot
be handed one. Tag parsing is defensive anyway: one unparseable tag must not
take the whole check down with it.
Also: the check is rate-limited to once a day rather than every start, the
dialog gained a "Skip this version" that survives a restart, the thread is now
joined on close, and the About dialog says SageTube. The three binary updaters
-- yt-dlp, Deno, ffmpeg -- legitimately track their own upstreams and are
deliberately untouched; ytsage_app_update's docstring says so, because "update"
is an overloaded word in this codebase.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The package version had been left at the scaffolded 0.1.0 while the released
tag was v5.4.0. Nothing in the code chose 0.1.0 -- it was npm-init's equivalent,
never updated -- and it made every version comparison meaningless: the window
title, the About dialog and the update check all read it.
ConfigManager merged a stored config over the defaults with dict.update(),
which is shallow. A config written by an older build carries partial "player"
and "feed" objects, and a shallow update replaces the whole nested default with
the partial one, so keys added since came back missing. The `or 15` and
`or "auto"` fallbacks at the call sites were load-bearing because of it. The
merge is now recursive, and keys present only in the stored file are kept so a
downgrade cannot destroy settings.
Configs now carry a config_version. A file written before 5.4.0 -- including
one inherited from an upstream YTSage install -- has its stored
check_app_updates cleared once, because that setting used to point at PyPI's
`ytsage` package and oop7/YTSage's releases, neither of which is this program.
The settings tab read a missing check_app_updates as enabled and persisted that
reading on OK, so merely opening Custom Options turned the checker back on.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Building, upstream tracking and YTSage's own README moved to the wiki; README
points there.
The history is not rewritten, unlike every other repository here. This one
still merges from oop7/YTSage and a rewrite would change every commit id, which
costs more than a clean history is worth in a fork. .github/ and
readme-translations/ are upstream's files and stay where upstream put them.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Homer
