Them being in the overlay controller required them to be static, which
forces the compiler to give them a memory location at compile time. They
also don't really fit into the overlaycontroller.
The old version waited for a set amount of time (2 seconds) before
checking and then exiting. The new logic checks every second for up to 7
seconds. This should deal with variances in startup time across
machines. On my machine the desktop overlay seems to be initialized
after 2-3 seconds.
Fixes
https://github.com/OpenVR-Advanced-Settings/OpenVR-AdvancedSettings/issues/63
Removes the ability for end users to change logging settings. This was
deemed unnecessary because logging is a developer troubleshooting tool
and users being able to change it could almost only introduce problems.
People who want to change the logging settings can compile from source.
The 2 seconds is arbitrarily chosen. The problem seems to be that
Valve's own overlays are not being initialized before custom overlays,
leading to the Desktop overlay sometimes not being present when our
overlay initializes.
The second if statement inside the first one checking a second time is
definitely a little disgusting, but for now it gets the job done.
The 4.0 default was arbitrarily chosen by whoever decided to initialize
the variable in the class.
Since the OpenVR initialization code is no longer inside the
overlaycontroller there is no need for it to be available during object
construction. Therefore it doesn't make sense to have it be an object.
The shutdown call was completely removed because shutdowns are handled
by an OpenVR event that calls the relevant shutdown code. The destructor
for the old object was never called, removing it adds clarity to the
actual operation of the code and also removes unnecessary code.
The old shutdown code did not correctly exit but instead just
continuously called the QApplication notify function until it
SEGFAULTed. This led to crashes whenever SteamVR was shut down. Users
with Windows Error Reporting turned on would get a 100+ MB crashdump
file along with increased shutdown time for SteamVR.
For clarity the shutdown functions should probably be collected in a
single function.
clang-tidy is currently set up with errors for all checks that do not
warn. This is to prevent regressing the repo.
Gradually these warnings should be removed in order to prevent further
regresssions in warnings that do not have many occurences.
`tval.tv_usec` is a (signed) `long` and `ssPrec->m_offset` is an
`unsigned int`. Clang warns upon realizing that we are attempting to
divide a signed `long` with an `unsigned int`.
Casting the `long` to `unsigned long` makes both operands `unsigned` and
allows for auto conversion to `size_t`.
The `size_t` cast is to make it apparent that the cast is correct.
The IPolicyConfig file is a third party file and since it describes a
third party API we can't make changes to it. Adding it via the system
includes disables warnings on most compilers.
The _var was chosen because it conflicts with the naming convention in
the rest of the file.
Fixing the error was easier than getting GCC to ignore errors in one
file, or using a pragma to mute the warning.
Introducing a GCC pragma would require disabling C4068 on MSVC (unknown
pragma).
Encapsulating compiler specific options in a separate file
makes it easier to reason about what is going to happen on
a specific qmake call.
clang-msvc and msvc are separate because they don't support
exactly the same compiler switches.
There are still many elements remaining from the old input system.
These have not been removed due to time constraints. They should
be removed as part of a general clean up process.
The PTT profile buttons have been replaced with a text telling users
to use the SteamVR bindings interface. This is for people who don't
read changelogs.
The old system used the old OpenVR input system which is
incompatible with the new IVRInput.
This change broke the AccessbilityTabController, but
that is to be removed because the functionality can be
found elsewhere.
Vasily