List of known issues that are not clearly identified as DETONATOR 2D bugs (yet).
Supporting threads using pthread requires some more complicated steps in the Emscripten build AND in the game deployment.
- Turn on the pthread support in the Emscripten build in emscripten/CMakeLists.txt
target_compile_options(GameEngine PRIVATE -pthread)
...
target_link_options(GameEngine PRIVATE -pthread)
...
install(FILES "${CMAKE_CURRENT_BINARY_DIR}/GameEngine.worker.js" DESTINATION "${CMAKE_CURRENT_LIST_DIR}/../editor/dist")
- When you deploy your game to a web server the host must turn on HTTP Cross-Origin policy flags in order to enable SharedArrayBuffer.
https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/SharedArrayBuffer
This can be done with a .htaccess file. See sample below
Header set Access-Control-Allow-Origin "https://your-domain.com"
Header set Cross-Origin-Embedder-Policy "require-corp"
Header set Cross-Origin-Resource-Policy "same-site"
Header set Cross-Origin-Opener-Policy "same-origin"
Header set Access-Control-Allow-Headers "range"
- The threads must be launched by the main thread from the browser's main event loop. In other words if your code does:
auto thread = std.:thread(...);
thread.join();
You're deadlocking forever since the main thread (the currently) calling thread must return to the browser's event loop in order to launch the thread, but joining the thread will block the calling thread!
The way to work around this is to use the Emscripten -sPTHREAD_POOL_SIZE=n build parameter which adds
code to initialize N number of threads on the application launch so that they're ready to go.
Pulseaudio audio issue that happens occasionally. Pulseaudio sound is crackling a lot and several buffer
underruns are reported during the stream playback. Increasing audio buffer size or the audio device poll frequency makes no difference.
Audio test cases might hang and never finish.
Issue started happening after some distro level system update.
Pulseaudio version 16.1
Current workaround is to try to restart pulseaudio.
$ pulseaudio -k
$ pulseaudio --start
Random glib failure (seen when running unit tests such as unit_test_gui and unit_test_ipc)
[enska@arch build]$ ./unit_test_gui
(process:17817): GLib-CRITICAL **: 13:03:37.700: g_hash_table_lookup: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_insert_internal: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_lookup: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_insert_internal: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_lookup: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_insert_internal: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_lookup: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_insert_internal: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_lookup: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_insert_internal: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_lookup: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_insert_internal: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_lookup: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_insert_internal: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_lookup: assertion 'hash_table != NULL' failed
(process:17817): GLib-CRITICAL **: 13:03:37.701: g_hash_table_insert_internal: assertion 'hash_table != NULL' failed
**
GLib-GObject:ERROR:../glib/gobject/gtype.c:2876:g_type_register_static: assertion failed: (static_quark_type_flags)
Bail out! GLib-GObject:ERROR:../glib/gobject/gtype.c:2876:g_type_register_static: assertion failed: (static_quark_type_flags)
Aborted (core dumped)
This smells like GIANT FUCKING DOOKIE regarding some global state and associated initialization order that is undefined and randomly starts failing.
The "fix" is to try to shuffle some target_link_libraries in the CMakeList.txt around to see if the issue resolves. The real fix would be to remove amateurs from ever contributing any fucking code anywhere in the world. If you can't do better than this just do everyone a fucking favour and don't ever write any publicly released code. Thank you!
Issues with HTTP 'Content-Type' header being set incorrectly.
- Python's http module fails to identify and set the Content-Type properly on .js GET
- Manually adding 'Content-Type' header in the 'end_headers' method adds a second 'Content-Type' header.
- Chrome gets confused about what is the actual ContentType and conflicting errors appear.
- Using threads (inclusion of .worker.js ?) causes the Content-Type issue to appear
- Failures to load the JS code in the browser. Errors (from Chromium) include:
Refused to execute script from 'http://localhost:8000/GameEngine.js' because its MIME type ('text/plain') is not executable.
self.onmessage @ GameEngine.worker.js:1
worker.js onmessage() captured an uncaught exception: NetworkError: Failed to execute 'importScripts' on 'WorkerGlobalScope': The script at 'http://localhost:8000/GameEngine.js' failed to load.
threadPrintErr @ GameEngine.worker.js:1
self.onmessage @ GameEngine.worker.js:1
GameEngine.worker.js:1 Error: Failed to execute 'importScripts' on 'WorkerGlobalScope': The script at 'http://localhost:8000/GameEngine.js' failed to load.
at self.onmessage (GameEngine.worker.js:1:1386)
Refused to execute script from 'http://localhost:8000/GameEngine.js' because its MIME type ('text/plain') is not executable.
self.onmessage @ GameEngine.worker.js:1
- Buttons when added to a dialog can have 'auto default' property which creates surprising behaviour when 'enter' is pressed. The button will capture the 'enter' and accept the dialog.
- Designer often borks and fails to show the auto default properly. DON'T TRUST THE DESIGNER!
- Sometimes bug happens when there's no autoDefault property in the .xml. Make sure to ADD autoDefault with false.
- Check the generated .xml for the auto default for the real value.
<widget class="QPushButton" name="btnCancel">
<property name="text">
<string>Cancel</string>
</property>
<property name="autoDefault">
<bool>false</bool>
</property>
</widget>
- This is about the way the QComboBox absolutely wants to resize itself according to it's content. Intention is probably all good but in practice this will often cause weird layout issues and mis-alignment.
- There's no "don't adjust contents" policy, but looks like using "fixed" size policy helps?
void MainWindow::something()
{
DlgFoobar dlg(this);
dlg.show();
...
dlg.exec();
}
Calling show before exec makes the dialog lose modality. In other words the user can click on other controls in the mainwindow.
Comments on exec()
So calling exec() without show() being called first works as expected. The dialog remains modal and the user is restricted to only applying input on the dialog.
The problem is that using exec() is not adequate.
Consider a case where you wish to restore the dialog state to a previous state, i.e. populate the dialog and its data in some meaningful way.
To do this with smooth UX you really need to:
- Load and restore the dialog geometry before showing the dialog.
If you first show and then restore the resizing will cause ugly visual flicker and disturbance. - Show it after the geometry has been restored
- Then load any other state, including state that might fail (for example imagine a file is no longer available)
3.1 You need to have the dialog open so that you can open a QMessageBox properly using the dialog as the window parent in order to open the msg box properly.
3.2 Finally call exec() and block the user to the dialog.