clang searches for its installed headers relative to the path of the
program calling it, which doesn't work when the caller is a custom
program running from directory other than the system clang installation
directory. Add a hook to the compilation database support to let the
user inject extra flags into the database, which can be used to add
`-isystem $PATH_TO_INSTALLED_CLANG_HEADERS` to the recorded commands, so
that when the clang library loads the compilation database, it finds the
headers.
psyke83 reports that Raspbian Buster with gcc 8.3.0 rejects the
`-isystem=dir` form. The `=dir` form does not work on any gcc I can
test, so conditionally keeping it for other systems does not seem
worthwhile.
Requested-by: psyke83 <https://github.com/dxx-rebirth/dxx-rebirth/pull/453>
Commit 452d58a0b5 switched from passing a list of strings as the
right-hand side of `in` to passing a parenthesized string. The new form
causes a TypeError if the left hand side is None:
```
>>> None in ('1', '2')
False
>>> None in ('1')
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: 'in <string>' requires string as left operand, not NoneType
```
Add a trailing comma to turn the right hand side back into a list of
strings, as it was before the Raspberry Pi commit.
Fixes: 452d58a0b5 ("Update Raspberry Pi Mesa build for Buster & Pi 4B")
The new Raspberry Pi 4B revision has a minimum requirement of Raspbian
buster, and uses the Mesa V3D driver by default; in fact, the legacy
Broadcom drivers are no longer compatible with its VideoCore VC6 chip, so
building against Mesa is mandatory for this Pi revision.
Unfortunately, the Debian buster version of the Mesa packages
have intentionally dropped the GLESv1 headers, making building of GLESv1
software impractical (even though the VC4/V3D driver can still run GLESv1
applications).
The OpenGL driver in conjunction with SDL2* runs at full
speed on Raspberry Pi 3B & 4B, so it makes sense to modify the "mesa"
build to this configuration in light of these facts.
* SDL2 is highly desirable due to the availability of the KMSDRM driver, which is
the only way to run dxx-rebirth outside of X (we can no longer use
the dispmanx/rpi driver of SDL1).
When targeting Raspberry Pi with legacy drivers, the following error
is observed on Raspbian buster with gcc 8.3.0:
cc1plus: error: =/opt/vc/include: No such file or directory [-Werror=missing-include-dirs]
cc1plus: error: =/opt/vc/include/interface/vcos/pthreads: No such file or directory [-Werror=missing-include-dirs]
cc1plus: error: =/opt/vc/include/interface/vmcs_host/linux: No such file or directory [-Werror=missing-include-dirs]
The error can be resolved by removing the equals sign from these definitions.
On all platforms, libSDL redefined `main` in a way that interfered with
SConf tests. On some platforms, but not on Linux, libSDL2 does not
redefine `main`. Based on a remark in the SDL2 migration guide, and the
apparent success (on Linux) after removing the `#undef`, I removed use
of `#undef` when building for SDL2. As privately reported by zico,
libSDL2 on Windows still redefines `main`, so it still needs an explicit
`#undef` to let the SConf tests build correctly. For platforms where
`#undef` is not needed, it is harmless. Add this `#undef`
unconditionally, and update the inline comment.
Fixes: f491059ed7 ("Enable building with SDL2")
Reported-by: zicodxx <private>
If register_compile_target=0, `self.builddir` on DXXProgram was used but
not set.
```
AttributeError: 'D1XProgram' object has no attribute 'builddir':
File "/s/SConstruct", line 5263:
main(register_program)
File "/s/SConstruct", line 5218:
''.join(['%s:\n%s' % (d.program_message_prefix, d.init(substenv)) for d in dxx])
File "/s/SConstruct", line 4853:
self.register_program()
File "/s/SConstruct", line 4895:
exe_target = self.builddir.File(exe_target)
```
pkubaj proposed extending the if/elif tree. This patch instead switches
to a loop, which avoids repeating any strings and is cleaner to extend
when another platform shows up needing normalization.
Requested-by: pkubaj <https://github.com/dxx-rebirth/dxx-rebirth/pull/426>
clang treats -Wunused as a request for -Wunused-lambda-capture, which is
good, except that -Wunused-lambda-capture is overzealous. It should
warn for captures that are not used in the body. It should not, but
does, warn for captures which are used in the body when such captures
are not required by the standard. Since 4 of the supported gcc versions
require that capture, removing the capture is not an option at this
point. Add a test to suppress the clang warning in this case, and hope
that eventually clang will warn only for uses that are actually unused.
This removes the need for parts of <https://github.com/dxx-rebirth/dxx-rebirth/pull/426> (by pkubaj).
Commit 0e81c05 moved the offending line, but it was already wrong in
54d57a37c9.
Fixes: 0e81c0594e ("Move settings that affect the build to UserBuildSettings")
Users who do not install Boost cannot include <boost/config.hpp>.
SConstruct handled this by disabling the macro DXX_BOOST_FALLTHROUGH,
but did not override gcc's default-enabled (via -Wextra)
use of -Wimplicit-fallthrough=3, so users would get a fallthrough
warning and the build would fail. Adjust this area to explicitly probe
for -Wno-implicit-fallthrough when Boost.Config is not available.
Fixes: 063bf29225 ("Enable -Wimplicit-fallthrough=5; fix resulting breaks")
Reported-by: Jayman2000 <https://github.com/dxx-rebirth/dxx-rebirth/issues/417>
gcc-4.9 lacks std::cbegin, std::cend. Fortunately, nothing in the code
uses those, so remove tests for them.
Fixes: 5e434cbe95 ("Require availability of C++11 begin")
Remove the fallback to Boost.Begin. C++14 is now the minimum supported
standard, and any conforming C++14 compiler should have a working C++11
std::begin.
Processing sharepath in SConf is incorrect, because targets can share a
build directory (and therefore an SConf run), but not share a sharepath.
Move sharepath handling out of SConf. Move DXX_USE_SHAREPATH handling
from CGameArg to GameArg, since one game can be built with a sharepath
while the other is built without.
The comma was omitted to discourage putting anything after the last
element. This was fine when there were multiple elements, but with only
one value, the missing comma causes this not to be a tuple as intended.
Add a comma to fix the test. Add some additional comments on supported
C++ versions by gcc major version.
Recent code changes require use of return type deduction, which is only
standard in C++14 and later. Require C++14. In practice, this changes
nothing, because the minimum supported compiler versions were already
choosing C++14. This change rejects older compilers that were already
unsupported.
PHYSFS_init is not guaranteed to succeed. Using PHYSFS functions after
PHYSFS_init fails is likely to fail badly. On Windows, failure may take
the form of a crash in ntdll. Avoid this by exiting gracefully.
Creating an attribute on UserBuildSettings during SConf breaks
clean/help, which do not run configure tests. Remove that and add a
method to return the value the attribute would have had.
Reported-by: derhass <https://github.com/dxx-rebirth/dxx-rebirth/issues/412>
Fixes: 1be414217c ("Add build time flag to choose whether to enable ADL MIDI support")
User jammer1 reports that Python under MSYS2[1] reports a `sys.platform`
of `msys`. For the limited purposes Rebirth requires, this platform can
be treated as if it were `win32`. Add a mapping to apply this change
automatically, so that users do not need to set it on the command line.
Reported-by: jammer1 <https://github.com/dxx-rebirth/dxx-rebirth/issues/405>
[1]: http://www.msys2.org/
enable_build_failure_summary=1 assumed that the command to display was a
list. For internal pseudo-commands, such as write_pch_inclusion_file,
the command was a string, so calling ' '.join() was incorrect. When the
output is not a list, pass it through unchanged.
Add macro cf_assert ("control flow" assert) to hint to gcc that certain
conditions are impossible. Use it to avoid generating range checks for
situations that never happen. If the event did happen, the only
consequence would be truncated UI text, rather than a correctness
problem.
The prior design was meant to allow the hook to be called instead of the
normal function, with the hook chaining to the normal function and then
performing other work as needed. This flexibility is unnecessary, and
complicates improvements elsewhere. Replace it with a hook that is
called with the result of the regular StaticObject call.
Use of the comprehensions makes the code less readable and harder to
modify. The performance benefit, if any, is in the noise. Switch to
an unpacked form.
This commit enables Rebirth to build with SDL2, but the result is not
perfect.
- SDL2 removed some sticky key support. Rebirth may behave differently
now in this area.
- SDL2 removed some key-repeat related support. Rebirth may behave
differently now in this area.
- SDL2 gained the ability to make a window fullscreen by sizing it to
the desktop instead of by changing the desktop resolution. Rebirth
uses this, and it mostly works.
- Resizing while in the automap does not notify the automap code, so
the view is wrong until the player switches out of automap mode and
back in.
- SDL2 changed how to enumerate available resolutions. Since
fitting the window to the desktop is generally more useful than
fitting the desktop to the window, I chose to drop support for
enumerating resolutions instead of porting to the new API. Users can
now enter an arbitrary window dimension and Rebirth will make an
attempt to use it.
- It might be useful to cap the window dimension at the desktop
dimension, but that is not done yet.
- Entering fullscreen mode through the Controls->Graphics submenu
failed to notify the relevant subsystems, causing the rendered
content not to rescale. For now, compile out the option to toggle
full screen through that menu. Toggling through Alt+Enter works
properly.
Despite these quirks, this is a substantial improvement over the prior
commit, where SDL2 cannot be used at all. The remaining issues can be
resolved in future work.
References: <https://github.com/dxx-rebirth/dxx-rebirth/issues/82>
Currently, DXX_USE_SDL_REDBOOK_AUDIO is an alias for !SDL2. However,
this patch enables interested users to readily patch out Redbook even
for SDL1.
References: <https://github.com/dxx-rebirth/dxx-rebirth/issues/82>
On Windows, SDL headers define `main` to `SDL_main` in addition to doing
it through `pkg-config`. The countering `#undef` must occur after the
headers are included. Commit 2a82207 changed the `#undef` to occur too
early. Linux works fine, but Windows broke.
Move the `#undef` back down. Switch to named format parameters to make
it easier to maintain.
Fixes: 2a82207e91 ("Add unit tests for valptridx parameter validation")