You can now run the test HTTP server by calling:
scons p=web serve
If you also wish to run the browser, call instead:
scons p=web run
The default listen port is 8060, but can be overriden via the env
variable GODOT_WEB_TEST_PORT which must be a valid integer.
On X11 we rely on ConfigureNotify event to update rendering context
window size, but we don't get such event when resizing
programmatically, only when done "manually" by resizing the window
etc.
Uses html-eslint for HTML file and eslint-plugin-html for inline
JavaScript.
Use HTML5 (not XHTML), remove CDATA and trailing slashes for self
closing tags.
Add format checks to CI.
Replace all TODO uses of `#warning` by proper TODO comments, and will open
matching bug reports to keep track of them.
We don't have a great track record fixing TODOs, but I'd wager we're even
worse for fixing these "TODO #warning" so we should prohibit this usage.
This makes it easier to set up, as you always have Python installed
when building Godot. On the other hand, you don't always have Node.js
+ npm installed (and you may not want to spend time running `npm install`).
Co-authored-by: Fabio Alessandrelli <fabio.alessandrelli@gmail.com>
This makes the setting easier to find, as research has found there are
numerous use cases to limiting FPS. This also improves documentation
related to the Engine property and project setting.
The project setting also works in projects exported in release mode,
so its location in the `debug/` section was misleading.
- `LIBC_FILEIO_ENABLED` wasn't defined anywhere, even in _other platforms_.
- `NO_NETWORK` is also never defined. It probably isn't enough anyway to
disable network APIs in the current codebase.
- `UNIX_SOCKET_UNAVAILABLE` is never defined in this code but used by some
other platforms, clarify that.
- `NO_STATVFS` can be removed as Android supports it since API level 19,
which is our current min SDK level. It's also only used for
`DirAccessUnix::get_space_left()` which is anyway overridden by
`DirAccessJAndroid::get_space_left()` so it shouldn't make a difference.
* Fixed documentation for `DirAccess.get_space_left()`.
- `NO_FCNTL` is likely also a remnant of early Android days, in current NDK
r23 it seems to be available. Also cleaned up unused `fcntl.h` includes.
- `NO_ALLOCA` is never defined, and we use alloca in many places now.
Android was the last platform to still attempt to disable RTTI (for binary
size), but both the Android editor and now the ICU library used by templates
need RTTI.
There could still be the possibility to support this for non-ICU template
builds (i.e. without the TextServerAdvanced module), but since this isn't one
of the build configurations we test regularly it's pretty risky to keep this
option only for that specific use case. And our code is already littered with
`dynamic_cast`s which weren't guarded with `!defined(NO_SAFE_CAST)`.
Fixes#66413.
DisplayServerX11 tracks some internal state about whether the window is
fullscreen, minimized, maximized or none. This commit queries the
display server when the window changes, so that this internal state can
be correctly updated.
Implements https://github.com/godotengine/godot-proposals/issues/3371.
New `target` presets
====================
The `tools` option is removed and `target` changes to use three new presets,
which match the builds users are familiar with. These targets control the
default optimization level and enable editor-specific and debugging code:
- `editor`: Replaces `tools=yes target=release_debug`.
* Defines: `TOOLS_ENABLED`, `DEBUG_ENABLED`, `-O2`/`/O2`
- `template_debug`: Replaces `tools=no target=release_debug`.
* Defines: `DEBUG_ENABLED`, `-O2`/`/O2`
- `template_release`: Replaces `tools=no target=release`.
* Defines: `-O3`/`/O2`
New `dev_build` option
======================
The previous `target=debug` is now replaced by a separate `dev_build=yes`
option, which can be used in combination with either of the three targets,
and changes the following:
- `dev_build`: Defines `DEV_ENABLED`, disables optimization (`-O0`/`/0d`),
enables generating debug symbols, does not define `NDEBUG` so `assert()`
works in thirdparty libraries, adds a `.dev` suffix to the binary name.
Note: Unlike previously, `dev_build` defaults to off so that users who
compile Godot from source get an optimized and small build by default.
Engine contributors should now set `dev_build=yes` in their build scripts or
IDE configuration manually.
Changed binary names
====================
The name of generated binaries and object files are changed too, to follow
this format:
`godot.<platform>.<target>[.dev][.double].<arch>[.<extra_suffix>][.<ext>]`
For example:
- `godot.linuxbsd.editor.dev.arm64`
- `godot.windows.template_release.double.x86_64.mono.exe`
Be sure to update your links/scripts/IDE config accordingly.
More flexible `optimize` and `debug_symbols` options
====================================================
The optimization level and whether to generate debug symbols can be further
specified with the `optimize` and `debug_symbols` options. So the default
values listed above for the various `target` and `dev_build` combinations
are indicative and can be replaced when compiling, e.g.:
`scons p=linuxbsd target=template_debug dev_build=yes optimize=debug`
will make a "debug" export template with dev-only code enabled, `-Og`
optimization level for GCC/Clang, and debug symbols. Perfect for debugging
complex crashes at runtime in an exported project.
- `_DEBUG` is MSVC specific so it didn't make much sense to define for
Android and iOS builds.
- iOS was the only platform to define `DEBUG`. We don't use it anywhere
outside thirdparty code, which we usually don't intend to debug, so it
seems better to be consistent with other platforms.
- Consistently define `NDEBUG` to disable assert behavior in both `release`
and `release_debug` targets. This used to be set for `release` for all
platforms, and `release_debug` for Android and iOS only.
- Due to the above, I removed the only use we made of `assert()` in Godot
code, which was only implemented for Unix anyway, should have been
`DEV_ENABLED`, and is in PoolAllocator which we don't actually use.
- The denoise and recast modules keep defining `NDEBUG` even for the `debug`
target as we don't want OIDN and Embree asserting all over the place.
- `-fomit-frame-pointer` is included automatically by both GCC and
Clang in `-O1` and above.
- `-ftree-vectorize` is included automatically by GCC in `-O2` and
beyond, and seems always enabled by Clang.
Closes#66296. See that issue for a detailed investigation.
Instead of updating all viewports, then blitting all viewports
to the backbuffer, then swapping all buffers, we run through
all viewports and render, blit, and swap backbuffer before
going to the next viewport.
- Don't warn about minimized/maximized modes not being available.
- Blender and FBX export both depend on running thirdparty applications,
which can't be done (easily at least) for Web and Android editors.
- Editor theme complained about not being able to retrieve texture data
for an icon. It was only used once so instead of flipping at runtime,
let's just add a flipped icon.
Part of #65702.
Adds support for LTO on macOS and Android. We don't have much experience
with LTO on these platforms so for now we keep it disabled by default
even when `production=yes` is set.
Similarly for iOS where we ship object files for the user to link in
Xcode so LTO makes builds extremely slow to link.
`production=yes` defaults to full LTO.
ThinLTO is much faster for LLVM-based compilers but seems to produce
bigger binaries (at least for the Web platform).
Removes separate `Command` key (use `Meta` instead).
Adds an event flag to automatically remap `Command` <-> `Control` (cannot be set alongside `Control` or `Meta`).
Fixes a crash due to an Xlib error, as well as ensures that Godot
holds the correct size of the window after window modes have been
applied, before exiting the DisplayServerX11 constructor. This ensures
the bootsplash will be displayed with the correct dimensions.
Fixes#65320
The DisplayServerX11 constructor processes pending events shortly
after constructing the main window. However, it discards pending events
on the event queue that it is not interested in. This results in these
events never making it to the main events thread and as a result are
never processed. We need to save the events we don't handle in
DisplayServerX11 so that they can be resent for later handling by the
events thread.
Attempts to construct an X11 window in an initial state of
minimized/maximized would fail due to the window being unmapped.
We simply check for failed mode changes during an unmap and reapply
them if necessary.
Update export names (web[_dlink]_[release|debug].zip).
The Build with dynamic linking is broken due to high number of imports
in output wasm (likely emscripten regression issue 15487).
Attempts to set a Godot window to fullscreen prior to the window being
mapped would silently fail. This commit uses
_window_fullscreen_check to test if a window had been set to fullscreen
while unmapped, and if so, resets it to fullscreen once the window has
been successfully mapped.
Fixes#54065
Replace a bisect with a single multiplication when calling
screen_get_dpi() in Javascript
Tested the value of
window.matchMedia(`(resolution:${(window.devicePixelRatio*96).toFixed(100)}dpi)`).matches
which is true except for values that cause a lot of rounding errors
(e.g. dpr : 0.3 => resolution: 28.799999999999997dpi)
Even in these cases the value matches the result of the previous
`findDPI()` method.
See also:
6cff589b5b (r81273660)
The issue was causing by a bug within the logic for `FileAccessFilesystemJAndroid#eof_reached()` causing that value to remain false after the eof was reached.
This in turn caused an infinite loop in the file scanner preventing the project's content from showing up.
This lets the engine and projects use the same color codes
in the terminal on all platforms. This includes support for
`print_rich()` (GDScript) and `print_line_rich()` (C++).
This was removed in #63481, and we confirmed that it's better like this,
but we add back the possibility to strip CR as an option, to optionally
restore the previous behavior.
For performance this is done directly in `String::parse_utf8`.
Also fixes Android `FileAccess::get_line()` as this one _should_ strip CR.
Supersedes #63717.
- Accelerate common path used to check the storage scope for a given path
- Update the logic for the `get_as_text()` method - previous logic loads the content of a text file one byte at a time
Previously it was used only on debug builds, but recent
additions to variant_call.cpp have made that unit too
large even in release_debug.
It was originally set to debug only because using it
breaks mingw, so this only reinstates it for MSVC where
it shouldn't have any drawbacks (famous last words).
Test-Information:
Builds for me with
scons -j 40 platform=windows target=release_debug
and my game then runs from it.
Implement built-in classes Vector4, Vector4i and Projection.
* Two versions of Vector4 (float and integer).
* A Projection class, which is a 4x4 matrix specialized in projection types.
These types have been requested for a long time, but given they were very corner case they were not added before.
Because in Godot 4, reimplementing parts of the rendering engine is now possible, access to these types (heavily used by the rendering code) becomes a necessity.
**Q**: Why Projection and not Matrix4?
**A**: Godot does not use Matrix2, Matrix3, Matrix4x3, etc. naming convention because, within the engine, these types always have a *purpose*. As such, Godot names them: Transform2D, Transform3D or Basis. In this case, this 4x4 matrix is _always_ used as a _Projection_, hence the naming.
The new option is `linker` and lets the user specify the argument to
the`-fuse_ld=` linker flag directly. The supported options are:
- `default`: No change, typically uses GNU ld (bfd) unless the user or
distro picked a different default `/usr/bin/ld`.
- `bfd`: GNU ld from binutils
- `gold`: GNU gold from binutils
- `lld`: lld from LLVM
- `mold`: mold, an extremely fast modern linker, not (yet) intended for
use in production but great for development speed. Provided by distro
`mold` package or needs to be compiled from source and installed to
`/usr` otherwise.
Removes the `use_lld=yes` option, and make lld actually usable with GCC
too.
Not all the above are compatible or recommend for LTO, we recommend
using GNU ld with GCC LTO, or lld with LLVM ThinLTO.
Now the `linuxbsd` platform can be built headlessly (e.g. without X11
development libraries).
I also cleaned up some weird (old?) usages of the `env` variable which
seem to make no difference and are used nowhere else.
This was done by refactoring directory and file access handling for the Android platform so that any general filesystem access type go through the Android layer.
This allows us to validate whether the access is unrestricted, or whether it falls under scoped storage and thus act appropriately.
Instead of reusing the custom_template/ prefix which is actually only
used for the prebuilt APK workflow, we add a new custom_build/ prefix.
This is a slight compat breakage (users will have to redo their config)
but enables us to group the Min SDK and Target SDK options where they make
sense, and avoid reusing the previously hardcoded Target SDK 30 from
Godot 3.4.
Those two options are now strings instead of integers so that we can keep
them empty by default, and show their default value using a placeholder.
So some validation has been added to make sure they are proper ints.
The upper bound on Target SDK was also removed as it's a common use case
to use it to try to target newer released SDKs. But we warn the user that
this wasn't validated by us.
The export info dialog is now exclusive so that when it doesn't auto-close,
i.e. when it errors, you don't close it by mistake by clicking outside.
Fixes#62465.
- Using a bucketized approach to select the editor scale in order to avoid too high values
- Add default app dimensions: used on Android devices with free floating app windows to set the default app frame
- Add ability to launch the Game window in an adjacent frame when in multi window mode
Increases the size of the wasm by around 3% (~300-350 KiB).
This enables using the Crypto object for hashing, signing and encryption,
and therefore reduces the gap between the features of the HTML5 platform
and other platforms.
Closes https://github.com/godotengine/godot-proposals/issues/3574.
This is in line with what's reported by the `OS` class on GNU/Linux, and
is required by the extension exporter to identify the correct library.
For BSD, we should either finish splitting the platform (into
platform/bsd) or register a separate exporter with OS name BSD and
proper templates detection.
* Replace case-by-case extraction with PNAME & GNAME
* Fix group handling when group hint begins with property name
* Exclude properties that are PROPERTY_USAGE_NO_EDITOR
* Extract missing ADD_ARRAY*, ADD_SUBGROUP* macros
* Map is unnecessary and inefficient in almost every case.
* Replaced by the new HashMap.
* Renamed Map to RBMap and Set to RBSet for cases that still make sense
(order matters) but use is discouraged.
There were very few cases where replacing by HashMap was undesired because
keeping the key order was intended.
I tried to keep those (as RBMap) as much as possible, but might have missed
some. Review appreciated!
popup no longer tries to close itself a second time
popup no longer closes after having been reopened
fixed bug in RenameDialog not calling base (by inspection)
fixes#59181fixes#60921
reverts #59287
Adds a new, cleaned up, HashMap implementation.
* Uses Robin Hood Hashing (https://en.wikipedia.org/wiki/Hash_table#Robin_Hood_hashing).
* Keeps elements in a double linked list for simpler, ordered, iteration.
* Allows keeping iterators for later use in removal (Unlike Map<>, it does not do much
for performance vs keeping the key, but helps replace old code).
* Uses a more modern C++ iterator API, deprecates the old one.
* Supports custom allocator (in case there is a wish to use a paged one).
This class aims to unify all the associative template usage and replace it by this one:
* Map<> (whereas key order does not matter, which is 99% of cases)
* HashMap<>
* OrderedHashMap<>
* OAHashMap<>
fixed and simplified gl_manager_windows
swap buffers now called for all windows
fixed missing pixel format setting in additional windows
this makes them work in OpenGL contexts
changed verbose error printing to write once
this error message happens very frequently while opengl3 is not finished
removed dead code no longer needed after changes
fixed comments that were misinformation
window messages during window creation now handled
these were previously discarded
messages now tunnel the required context
changed failure to create opengl3 window on windows to be more fatal
marked a problem with pen code
conditional compilation of vulkan and opengl3 on windows fixed
windows debug builds now show messages on debug console also
rendering driver selection box now shows only compiled drivers
marked some problematic code
thanks to akien-mga for patiently rewriting my style mistakes
Adds the is_process_running function to the native OS class and exposes it to script.
This is implemented on Windows and Unix platforms. A stub is provided for other platforms that do not support this function.
Documentation is updated to reflect new API function.
These typedefs don't save much typing compared to the full `Ref<Resource>`
and `Ref<RefCounted>`, yet they sometimes introduce confusion among
new contributors.
Didn't commit all the changes where it wants to initialize a struct
with `{}`. Should be reviewed in a separate PR.
Option `IgnoreArrays` enabled for now to be conservative, can be
disabled to see if it proposes more useful changes.
Also fixed manually a handful of other missing initializations / moved
some from constructors.
It's not needed on most distros as those are found in standard lib
and include paths, but on NixOS they're all in non-standard prefixes,
so we need to rely on information provided by pkg-config.
Fixes#59913.
Co-authored-by: David Lewis <davidalewis00@gmail.com>
Windows export process:
Limit size of executable with embedded PCK to 4 GB.
Use "rcedit" before embedding PCK.
Capture and process "rcedit" errors.
Windows, Linux:
Add support for PCK loading from executable "pck" section.
The multithreaded importer often causes system-wide hangs when importing more than 20 files at a time.
Running the editor at normal priority allows other applications on the system to be responsive during long imports.
These set of changes focus primarily on getting the core logic and overall Godot Editor UI and functionality up and running natively on Android devices.
UI tweaks / cleanup / polish, as well configuration for Android specific functionality / restrictions will be addressed in follow-up PRs iteratively based on feedback.
Co-authored-by: thebestnom <shoval.arad@gmail.com>
* DisplayServerOSX::mouse_get_position returns wrong x axis data in
multi-monitor case, which makes mouse_process_popups send
WINDOW_EVENT_CLOSE_REQUEST, then eventually make PopupMenu close on
mouse down without activating item.
`DirAccess *` needs to be deleted manually, and this is often forgotten
especially when doing early returns with `ERR_FAIL_COND`.
`DirAccessRef` is deleted automatically when it goes out of scope.
Co-authored-by: bruvzg <7645683+bruvzg@users.noreply.github.com>
* Previous "virtual" classes (which can't be instantiated) are not corretly named "abstract".
* Added a new "virtual" category for classes, they can't be instantiated from the editor, but can be inherited from script and extensions.
* Converted a large amount of classes from "abstract" to "virtual" where it makes sense.
Most classes that make sense have been converted. Missing:
* Physics servers
* VideoStream
* Script* classes.
which will go in a separate PR due to the complexity involved.
It has been disabled in `master` since one year (#45852) and our plan
is for Bullet, and possibly other thirdparty physics engines, to be
implemented via GDExtension so that they can be selected by the users
who need them.
* Very old macros from the time Godot was created.
* Limited arguments to 5 (then later changed to 8) in many places.
* They were replaced by C++11 Variadic Templates.
* Renamed methods that take argument pointers to have a "p" suffix. This was used in some places and not in others, so made it standard.
* Also added a dereference check for Variant*. Helped catch a couple of bugs.
This has been superseded by GDExtension so this code is no longer useful
nor usable.
There's still some GDNative-related stuff in platform export code which
needs to be adapted for GDExtension (e.g. to include GDExtension libraries
in exports).
* Its not and will not be used for streaming.
* Streaming will be implemented in 4.1 and it will work different.
* It makes more sense to be called CompressedTexture since it imports and compresses texture files.
Add localizable string (Dictionary<Lang Code, String>) property editor and property hint.
Add localized "app name" property to the project settings.
Add localized permission and copyright properties to the macOS and iOS export settings.
Remove some duplicated ("app name") and deprecated ("info") macOS and iOS export properties.
This method can be used to get the CPU model name.
It can be used in conjunction with
`RenderingServer.get_video_adapter_name()` and
`RenderingServer.get_video_adapter_vendor()` for annotating benchmarks
and automatic graphics quality configuration.
Removes some unnecessary includes from `editor_node.h`, and instead add
those where they're used.
Removes unnecessary `editor_node.h` includes in various editor classes.
Renames `dynamicfont` to `dynamic_font` in a couple files.
Misc cleanup while jumping through that rabbit hole.
My Mac was using 20% cpu again, which was related to the Javascript
Export plugin.
I had however no export templates setup in the project so this is more
of a stopgap fix.
Implemented via `BCryptGenRandom` on Windows.
Implemented via `getentropy` syscall when available.
Implemented via `/dev/urandom` device as a fallback.
The `/dev/urandom` fallback can be disabled via the `NO_URANDOM` build
flag.
Note: The HTML5 version relies on emscripten file system urandom
device which itself uses the Crypto API when available or the plain
old not crypto-safe `Math.random()` otherwise.
Restore get_entropy.
Using codespell 2.2-dev from current git.
Added `misc/scripts/codespell.sh` to make it easier to run it once in a
while and update the skip and ignore lists.
All iOS devices since the iPhone 5S support ARMv8 (64-bit).
The last iOS version supported on ARMv7 devices is 10.x, which is
too old to run Godot 4.0 projects since the minimum supported
iOS version is 11.0.
Both are recognized by Xcode and equivalent, but osxcross issues a
warning for the latter:
```
osxcross: warning: '-mmacosx-version-min=' (11.0.0 != 11.00)
```
Modify the create_sub_window method to set an icon for all sub windows, setting the icon to the same icon as the main window.
Co-Authored-By: Rémi Verschelde <rverschelde@gmail.com>
Allows detecting when a new version of the progressive web app service
worker is waiting (i.e. an update is pending), along a function to force
the update and reload all clients.
Use an offline first approach, where we prefer the cached version over
the network one.
This forces games using PWA to always re-export the project and not just
the PCK, so that the service worker version gets updated correctly, and
the end-user cache is correctly cleared on update.
Replace WM_MOVE and WM_SIZE message handling with WM_POSCHANGED instead.
This is for multiple reasons:
1) Microsoft suggest using WM_POSCHANGED is more efficient
2) RectChanged callback is only called once for most window operations
3) Simplifies message handling code
- Adds the parameters for supported Meta devices, which is required to access some device specific capabilities
- Remove the 'com.samsung.android.vr.application.mode' metadata
Running `npm run serve` in `platform/javascript/` will start
a web server that uses `bin/` as its root folder. This means you
can extract a compiled web editor ZIP in this folder and immediately
be able to test the web editor.
Headers required for the use of SharedArrayBuffer are automatically
added to every response.
A common source of errors is to call functions (such as round()) expecting them to work in place, but them actually being designed only to return the processed value. Not using the return value in this case in indicative of a bug, and can be flagged as a warning by using the [[nodiscard]] attribute.
[macOS] Fix transient windows not working in the full-screen mode.
[macOS] Fix moving transient windows to the other screen than parent window.
[macOS] Fix popup menu switch on hover.
[macOS] Use content origin rect for windows position (to ensure `DS.mouse_get_position` is equal to `DS.window_get_position` + mouse position from the input events).
[macOS] Fix incorrect input coordinates, when external display with different scaling in connected/disconnected.
[macOS/Windows] Fix moving fullscreen windows between the screens.
Add auto refocusing of the parent window, when the focused transient window is closed.
Remove redundant `DS.mouse_get_absolute_position` function (returns mouse position in the screen coordinates, same as `DS.mouse_get_position`).
Merge "required" / "optional" icons into a single list.
Remove "generate_missing" and automatically rescale icons / loading screens that are missing or have incorrect size.
Print warning if icon or loading screen has incorrect size.
On the only platform where PVRTC is supported (iOS),
ETC2 generally supersedes PVRTC in every possible way. The increased
memory usage is not really a problem thanks to modern iOS' devices
processing power being higher than its Android counterparts.
Found via `codespell -q 3 -S ./thirdparty,*.po,./DONORS.md -L ackward,ang,ans,ba,beng,cas,childs,childrens,dof,doubleclick,expct,fave,findn,gird,hist,inh,inout,leapyear,lod,nd,numer,ois,ony,paket,ro,seeked,sinc,switchs,te,uint,varn,vew`
Each file in Godot has had multiple contributors who co-authored it over the
years, and the information of who was the original person to create that file
is not very relevant, especially when used so inconsistently.
`git blame` is a much better way to know who initially authored or later
modified a given chunk of code, and most IDEs now have good integration to
show this information.
Always build with the GUI subsystem.
Redirect stdout and stderr output to the parent process console.
Use CreateProcessW for blocking `execute` calls with piped stdout and stderr (prevent console windows for popping up when used with the GUI subsystem build, and have more consistent behavior with `create_process`).
Add `open_console` argument to the `execute` and `create_process` to open a new console window.
Remove `interface/editor/hide_console_window` editor setting.
Remove `Toggle System Console` menu option.
Remove `set_console_visible` and `is_console_visible` functions.
We prefer to prevent using chained assignment (`T a = b = c = T();`) as this
can lead to confusing code and subtle bugs.
According to https://en.wikipedia.org/wiki/Assignment_operator_(C%2B%2B), C++
allows any arbitrary return type, so this is standard compliant.
This could be re-assessed if/when we have an actual need for a behavior more
akin to that of the C++ STL, for now this PR simply changes a handful of
cases which were inconsistent with the rest of the codebase (`void` return
type was already the most common case prior to this commit).
Ensure better compatibility when emcc which may run some tools from
different paths (e.g. closure compiler).
This fixes externs include issues with modern emcc using the closure
compiler.
This actually makes sense(?), when running inside an iframe the active
element might be our canvas, while the iframe itself is not active in
the parent window. Since we consume the event, the iframe does not get
focused in Firefox (but does in Chromium-based browsers), so we must
always call focus to handle such occasions.
Note, the editor build requires the mbedtls module to be manually
enabled, as it is currently needed as a ResourceUID dependency.
This will need to be addressed in a separate PR.
In some conditions the events might be generated even when the `gamepad`
object is not accessible due to Security Context requirements.
This commit adds a check to avoid firing the handler in those cases.
Building `target=release` and `target=release_debug` builds with MinGW-GCC
errors when linking with LTO.
Since it's only needed for `target=debug` builds anyway (bigger objects), which
we don't build with LTO, this works around the issue.
Equivalent `-Wa,-mbig-obj` for GCC/Clang.
This started being needed to compile harfbuzz in `target=debug` with MinGW/GCC,
but there doesn't seem to be any drawback to enabling `/bigobj` (aside from
losing support for pre-VS 2005 linkers, which we don't support).
A window can be closed on the server side while processing results from
_NET_CLIENT_LIST, which causes BadWindow fatal errors by default in
XGetWindowProperty.
The only way to safely catch this case is to set an error handler to
ignore BadWindow errors while these commands are processed.
