Move Win32 async context ownership and trim canvas draw setup

This commit is contained in:
2026-06-16 08:49:31 +02:00
parent d5b137c9ff
commit 667589f1f6
7 changed files with 181 additions and 93 deletions

View File

@@ -18,6 +18,17 @@ agent or engineer to remove them without reconstructing context from chat.
## Reductions
- 2026-06-16: `DEBT-0003` was narrowed again. The Win32 async GL/context lock
state now lives in `src/platform_windows/windows_platform_services.cpp`
instead of `src/main.cpp` retained state, and `main.cpp` only seeds the
created `HDC`/`HGLRC` into that platform-owned helper during initialization
and context recreation; broader Win32 bootstrap/runtime ownership remains.
- 2026-06-16: `DEBT-0036` was narrowed again. `NodeCanvas` merged-path
per-plane merged-texture draw callback setup now routes through
`make_legacy_canvas_draw_merge_layer_texture_draw(...)` in
`src/legacy_canvas_draw_merge_services.h` instead of keeping that callback
setup inline in `NodeCanvas::draw()`; broader canvas draw orchestration and
retained GL resource ownership remain.
- 2026-06-16: `DEBT-0037` was narrowed again. `App::rec_loop()` in
`src/app.cpp` now delegates recording worker iteration orchestration into
`process_legacy_recording_worker_iteration(...)` in

View File

@@ -117,9 +117,9 @@ Current architecture mismatches that must be treated as real blockers:
checkerboard background setup now route through retained draw-merge helpers,
with the cache-to-screen checkerboard-plane callback setup also reduced and
the merged-path checkerboard background-plane callback plus per-plane
merged-texture draw plus the smoothing-mask face shader/draw pass plus
heightmap, current-mode, and grid-mode callback setup now routed through the
same retained helper family.
merged-texture draw callback plus the smoothing-mask face shader/draw pass
plus heightmap, current-mode, and grid-mode callback setup now routed
through the same retained helper family.
- `app_layout.cpp` and `app_dialogs.cpp` are still mixed shell/controller files
rather than thin composition/binding surfaces.
- `App`, `Canvas`, `Node`, retained workers, and platform entrypoints still use
@@ -135,13 +135,15 @@ Current architecture mismatches that must be treated as real blockers:
coordination flags now use `std::atomic` instead of unsynchronized globals,
while the main Win32 entrypoint now groups window/GL/task/VR state behind a
retained local state object instead of separate process-wide globals, the
canvas async worker now sits behind a named retained local worker-state
helper instead of a bare static accessor, the prepared-file worker and the
canvas async import/export/save/open worker now live under `AppRuntime`
instead of retained static app-events/canvas workers, and `App::rec_loop()`
now delegates worker-iteration orchestration into the retained recording
bridge even though that retained recording path still owns the worker-side
readback flow.
Win32 async GL/context lock state now lives under
`src/platform_windows/windows_platform_services.cpp` instead of `main.cpp`
retained state, the canvas async worker now sits behind a named retained
local worker-state helper instead of a bare static accessor, the
prepared-file worker and the canvas async import/export/save/open worker now
live under `AppRuntime` instead of retained static app-events/canvas
workers, and `App::rec_loop()` now delegates worker-iteration orchestration
into the retained recording bridge even though that retained recording path
still owns the worker-side readback flow.
- Modern C++23 usage exists in extracted components, especially `std::span`,
explicit result/status objects, and a few concepts, but the live app still
does not consistently express ownership, thread affinity, or renderer

View File

@@ -155,6 +155,10 @@ Current slice:
through `make_legacy_canvas_draw_merge_background_checkerboard_plane(...)`,
but the node still owns broader live layer traversal and renderer-state
sequencing.
- `NodeCanvas` merged-path per-plane merged-texture draw callback setup now
also routes through `make_legacy_canvas_draw_merge_layer_texture_draw(...)`,
but the node still owns broader live layer traversal and renderer-state
sequencing.
- `NodeCanvas` smoothing-mask face shader setup plus per-face draw execution
now also route through
`execute_legacy_canvas_draw_merge_smask_faces(...)`, but the node still owns
@@ -356,6 +360,10 @@ Current slice:
- `main.cpp` Win32 window handles, GL task/mutex state, splash-dialog state,
stylus timers, and VR worker state now sit behind one retained local state
object instead of separate file-scope globals
- Win32 async GL/context lock state now lives in
`src/platform_windows/windows_platform_services.cpp` instead of `main.cpp`
retained state, and `main.cpp` only seeds that platform-owned context handle
pair during initialization and context recreation
- prepared-file background work now runs through an `AppRuntime`-owned worker
queue instead of a retained static worker in `src/app_events.cpp`
- canvas async import/export/save/open background work now also runs through an

View File

@@ -494,6 +494,45 @@ inline void execute_legacy_canvas_draw_merge_layer_texture(
execution.unbind_layer_texture();
}
template <typename LayerMergeT, typename SamplerT, typename FacePlaneT, typename SetActiveTextureUnit, typename PlaneTransform>
[[nodiscard]] inline auto make_legacy_canvas_draw_merge_layer_texture_draw(
LayerMergeT* layer_merge,
SamplerT* sampler,
FacePlaneT* face_plane,
SetActiveTextureUnit set_active_texture_unit,
glm::mat4 proj,
glm::mat4 camera,
PlaneTransform plane_transform)
{
return [layer_merge, sampler, face_plane, set_active_texture_unit, proj, camera, plane_transform](int plane_index) {
const auto mvp = proj * camera *
plane_transform[plane_index] *
glm::translate(glm::vec3(0, 0, -1));
execute_legacy_canvas_draw_merge_layer_texture(
{
.mvp = mvp,
.texture_slot = 0,
.alpha = 1.f,
.highlight = false,
},
{
.bind_sampler = [sampler, set_active_texture_unit] {
sampler->bind(0);
set_active_texture_unit(0);
},
.bind_layer_texture = [layer_merge, plane_index] {
layer_merge->rtt(plane_index).bindTexture();
},
.draw = [face_plane] {
face_plane->draw_fill();
},
.unbind_layer_texture = [layer_merge, plane_index] {
layer_merge->rtt(plane_index).unbindTexture();
},
});
};
}
inline void execute_legacy_canvas_draw_merge_layer_composite(
bool is_temporary_erase,
bool is_temporary_paint,

View File

@@ -38,21 +38,22 @@
#define WM_USER_WAKEUP (WM_USER + 2)
LRESULT CALLBACK WndProc(HWND hWnd, UINT msg, WPARAM wp, LPARAM lp);
namespace pp::platform::windows {
void set_async_render_context(HDC hdc, HGLRC hrc);
void lock_async_render_context();
bool try_lock_async_render_context();
void unlock_async_render_context();
void swap_async_render_context();
}
struct RetainedState
{
HINSTANCE hInst{};
HWND hWnd{};
HDC hDC{};
HGLRC hRC{};
const wchar_t* className{};
bool keys[256]{};
std::mutex gl_mutex;
std::mutex async_mutex;
std::thread::id gl_thread{};
std::map<kKey, int> vkey_map;
wchar_t window_title[512]{};
std::jthread hmd_renderer;
int gl_count = 0;
std::deque<std::packaged_task<void()>> main_tasklist;
std::mutex main_task_mutex;
float timer_stylus = 0;
@@ -170,57 +171,22 @@ void destroy_window()
void async_lock()
{
//std::lock_guard<std::mutex> _lock(async_mutex);
auto& state = retained_state();
if (state.gl_count == 0 || state.gl_thread != std::this_thread::get_id())
{
state.gl_mutex.lock();
bool ret = wglMakeCurrent(state.hDC, state.hRC);
if (ret == false)
LOG("FAILED wglMakeCurrent: %s", GetLastErrorAsString().c_str());
state.gl_thread = std::this_thread::get_id();
//LOG("lock");
}
state.gl_count++;
//assert(ret == true);
pp::platform::windows::lock_async_render_context();
}
bool async_lock_try()
{
//std::lock_guard<std::mutex> _lock(async_mutex);
auto& state = retained_state();
if (state.gl_count == 0 || state.gl_thread != std::this_thread::get_id())
{
if (!state.gl_mutex.try_lock())
return false;
bool ret = wglMakeCurrent(state.hDC, state.hRC);
if (ret == false)
LOG("FAILED wglMakeCurrent: %s", GetLastErrorAsString().c_str());
state.gl_thread = std::this_thread::get_id();
//LOG("lock");
}
state.gl_count++;
//assert(ret == true);
return true;
return pp::platform::windows::try_lock_async_render_context();
}
void win32_async_swap()
{
SwapBuffers(retained_state().hDC);
//LOG("swap");
pp::platform::windows::swap_async_render_context();
}
void async_unlock()
{
//std::lock_guard<std::mutex> _lock(async_mutex);
auto& state = retained_state();
state.gl_count--;
if (state.gl_count == 0)
{
//LOG("unlock");
wglMakeCurrent(0, 0);
state.gl_mutex.unlock();
}
pp::platform::windows::unlock_async_render_context();
}
void win32_update_stylus(float dt)
@@ -866,11 +832,12 @@ int main(int argc, char** argv)
pfd.cDepthBits = 24;
pfd.iLayerType = PFD_MAIN_PLANE;
state.hDC = GetDC(state.hWnd);
int pxfmt = ChoosePixelFormat(state.hDC, &pfd);
SetPixelFormat(state.hDC, pxfmt, &pfd);
state.hRC = wglCreateContext(state.hDC); // Create OpenGL 2.1 or less
wglMakeCurrent(state.hDC, state.hRC);
HDC hDC = GetDC(state.hWnd);
int pxfmt = ChoosePixelFormat(hDC, &pfd);
SetPixelFormat(hDC, pxfmt, &pfd);
HGLRC hRC = wglCreateContext(hDC); // Create OpenGL 2.1 or less
wglMakeCurrent(hDC, hRC);
pp::platform::windows::set_async_render_context(hDC, hRC);
// Initialize extensions
if (!gladLoadGL())
@@ -878,7 +845,7 @@ int main(int argc, char** argv)
LOG("gladLoadGL() failed");
return 0;
}
if (!gladLoadWGL(state.hDC))
if (!gladLoadWGL(hDC))
{
LOG("gladLoadWGL() failed");
return 0;
@@ -912,18 +879,19 @@ int main(int argc, char** argv)
UINT numFormat;
wglMakeCurrent(NULL, NULL);
wglDeleteContext(state.hRC);
wglDeleteContext(hRC);
DestroyWindow(state.hWnd);
state.hWnd = CreateWindow(wc.lpszClassName, state.window_title, wnd_style, clientPos.x, clientPos.y,
(float)(clientRect.right - clientRect.left),
(float)(clientRect.bottom - clientRect.top), 0, 0, state.hInst, 0);
state.hDC = GetDC(state.hWnd);
wglChoosePixelFormatARB(state.hDC, wgl_config.pixel_format_attributes.data(), nullptr, 1, &pxfmt, &numFormat);
SetPixelFormat(state.hDC, pxfmt, &pfd);
state.hRC = wglCreateContextAttribsARB(state.hDC, NULL, wgl_config.context_attributes.data());
wglMakeCurrent(state.hDC, state.hRC);
hDC = GetDC(state.hWnd);
wglChoosePixelFormatARB(hDC, wgl_config.pixel_format_attributes.data(), nullptr, 1, &pxfmt, &numFormat);
SetPixelFormat(hDC, pxfmt, &pfd);
hRC = wglCreateContextAttribsARB(hDC, NULL, wgl_config.context_attributes.data());
wglMakeCurrent(hDC, hRC);
pp::platform::windows::set_async_render_context(hDC, hRC);
}
else
{

View File

@@ -385,32 +385,14 @@ void NodeCanvas::draw()
}),
});
const auto draw_merged_texture_plane = [&](int plane_index) {
pp::panopainter::execute_legacy_canvas_draw_merge_layer_texture(
{
.mvp = proj * camera *
m_canvas->m_plane_transform[plane_index] *
glm::translate(glm::vec3(0, 0, -1)),
.texture_slot = 0,
.alpha = 1.f,
.highlight = false,
},
{
.bind_sampler = [&] {
m_sampler.bind(0);
set_active_texture_unit(0);
},
.bind_layer_texture = [&] {
m_canvas->m_layers_merge.rtt(plane_index).bindTexture();
},
.draw = [&] {
m_face_plane.draw_fill();
},
.unbind_layer_texture = [&] {
m_canvas->m_layers_merge.rtt(plane_index).unbindTexture();
},
});
};
const auto draw_merged_texture_plane = pp::panopainter::make_legacy_canvas_draw_merge_layer_texture_draw(
&m_canvas->m_layers_merge,
&m_sampler,
&m_face_plane,
set_active_texture_unit,
proj,
camera,
m_canvas->m_plane_transform);
for (int plane_index = 0; plane_index < 6; plane_index++)
{

View File

@@ -26,6 +26,84 @@ void win32_update_stylus(float dt);
void win32_save_window_state();
bool win32_vr_start();
void win32_vr_stop();
std::string GetLastErrorAsString();
namespace pp::platform::windows {
namespace {
struct RetainedWin32AsyncRenderContextState final {
HDC hdc{};
HGLRC hrc{};
std::mutex gl_mutex;
std::thread::id gl_thread{};
int gl_count = 0;
};
[[nodiscard]] RetainedWin32AsyncRenderContextState& retained_win32_async_render_context_state()
{
static RetainedWin32AsyncRenderContextState state;
return state;
}
}
void set_async_render_context(HDC hdc, HGLRC hrc)
{
auto& state = retained_win32_async_render_context_state();
state.hdc = hdc;
state.hrc = hrc;
state.gl_thread = {};
state.gl_count = 0;
}
void lock_async_render_context()
{
auto& state = retained_win32_async_render_context_state();
if (state.gl_count == 0 || state.gl_thread != std::this_thread::get_id())
{
state.gl_mutex.lock();
const bool ret = wglMakeCurrent(state.hdc, state.hrc);
if (!ret)
LOG("FAILED wglMakeCurrent: %s", GetLastErrorAsString().c_str());
state.gl_thread = std::this_thread::get_id();
}
state.gl_count++;
}
bool try_lock_async_render_context()
{
auto& state = retained_win32_async_render_context_state();
if (state.gl_count == 0 || state.gl_thread != std::this_thread::get_id())
{
if (!state.gl_mutex.try_lock())
return false;
const bool ret = wglMakeCurrent(state.hdc, state.hrc);
if (!ret)
LOG("FAILED wglMakeCurrent: %s", GetLastErrorAsString().c_str());
state.gl_thread = std::this_thread::get_id();
}
state.gl_count++;
return true;
}
void unlock_async_render_context()
{
auto& state = retained_win32_async_render_context_state();
state.gl_count--;
if (state.gl_count == 0)
{
wglMakeCurrent(0, 0);
state.gl_mutex.unlock();
}
}
void swap_async_render_context()
{
SwapBuffers(retained_win32_async_render_context_state().hdc);
}
}
namespace {