Age | Commit message (Collapse) | Author | |
---|---|---|---|
2017-03-22 | compat/nan: support for mingw-w64 in a hacky way | Stanislaw Halik | |
2017-03-22 | compat/shm: decruft | Stanislaw Halik | |
2017-03-22 | compat/timer: accidental pragma once | Stanislaw Halik | |
2017-03-21 | compat/timer: move from header. simplify | Stanislaw Halik | |
It was getting inlined in each compilation unit. | |||
2017-03-20 | compat/sleep: guard against signed -> unsigned wraparound | Stanislaw Halik | |
2017-03-17 | compat/util: remove our make_unique custom impl | Stanislaw Halik | |
2017-02-27 | remove unfinished polish translation files | Stanislaw Halik | |
Polish speakers typically know English well enough and translations are awkward. | |||
2017-02-25 | contrib/util: add more unused attribute macros | Stanislaw Halik | |
2017-02-21 | modules: add .ts files | Stanislaw Halik | |
2016-12-31 | compat/timer: prevent races in initializing timer resolution | Stanislaw Halik | |
Also, - don't lose high bits with cast from long long to long on _WIN32. - reduce #define pollution | |||
2016-12-30 | compat: prevent deadlock with race | Stanislaw Halik | |
We can't depend on cvar getting notified only after "src" runs out of scope. Now, in case signal destroyed() runs first: - mtx locked - flag set to true - empty cvar notified Thus, doesn't hang. In case we wait first: - mtx locked - flag is false - cvar notification arrives Of course semaphore code always runs serially since they're covered by a mutex. We have all our bases covered. Previously the code never hung simply because the "curthread" condition was always true. I removed the "curthread" code paths since they don't add anything. Also rvalue references got used incorrectly. | |||
2016-12-30 | compat: remove empty translation unit | Stanislaw Halik | |
2016-12-16 | compat/util: implicitly include QDebug to every callee | Stanislaw Halik | |
2016-12-08 | fix build | Stanislaw Halik | |
Spotted-by: @MrSanchos | |||
2016-12-03 | compat: add asserts | Stanislaw Halik | |
2016-12-03 | fixup! tracker/{aruco,pt}: use RAII for QObject | Stanislaw Halik | |
2016-12-03 | compat: avoid namespace clashes | Stanislaw Halik | |
2016-11-29 | compat/run-in-thread: fix on msvc 2017 | Stanislaw Halik | |
2016-11-12 | compat: move from simple-mat.hpp to euler.hpp | Stanislaw Halik | |
2016-11-05 | delete empty .ts | Stanislaw Halik | |
2016-11-04 | modules: add autogenerated translation files | Stanislaw Halik | |
2016-11-03 | move simple-mat from logic to compat to simplify dependencies | Stanislaw Halik | |
2016-10-26 | compat/util: split out run-in-thread{,-async} | Stanislaw Halik | |
2016-10-26 | compat/util: introduce "iround" for round to integer | Stanislaw Halik | |
2016-10-26 | compat/util: move stuff around | Stanislaw Halik | |
2016-10-23 | compat/util: these type params are unused | Stanislaw Halik | |
2016-10-20 | compat/nan: fix isnan in presence of -ffast-math | Stanislaw Halik | |
The trick prevents GCC isnan builtin from getting called. | |||
2016-10-19 | compat, tracker/pt: add warn_unused_result compat macro | Stanislaw Halik | |
2016-10-19 | compat/util: fix conversion from value<t> for clamp | Stanislaw Halik | |
2016-09-21 | many: remove compat/pi-constant.hpp | Stanislaw Halik | |
With -D_USE_MATH_DEFINES MSVC defines the standard M_PI and friends. Since this preprocessor definition is now always passed as part of the build system for MSVC. We can use M_PI as if on a mission. | |||
2016-09-20 | compat/timer: no exports here | Stanisław Halik | |
2016-09-16 | compat, dinput: don't overdo COM logic | Stanislaw Halik | |
2016-09-16 | compat/util: replace with QSignalBlocker usages | Stanislaw Halik | |
2016-09-09 | compat/win32-com: switch to always apartment threaded | Stanislaw Halik | |
2016-09-08 | compat: remove empty sources | Stanislaw Halik | |
2016-08-23 | compat/util: move typical "clamp" function here | Stanislaw Halik | |
2016-08-23 | cmake: refactor boilerplate for projects | Stanislaw Halik | |
Don't rely on a macro that shouldn't be necessary. The information "opentrack_boilerplate" retrieves is now factored out to be easily available to projects themselves. opentrack_boilerplate can now also build executables. When appropriate, target properties are now concatenated rather than replaced. | |||
2016-08-23 | compat/util: add time-saving inhibit_qt_signals overloads | Stanislaw Halik | |
2016-08-17 | compat/win32-com: default to right COM type without caller specifying it | Stanislaw Halik | |
The main window always needs the apartment threaded context for OLE used by Qt. Default to apartment threaded before a QApplication is created, or if the thread is the QApplication thread. Use multithreaded COM otherwise. | |||
2016-08-17 | move options framework into its own library | Stanislaw Halik | |
- adjust usages - add support for QList signals and metatype | |||
2016-08-16 | compat/util: add robust way to block qt signals | Stanislaw Halik | |
2016-08-16 | compat/options: anon namespace classes are usable so move to detail | Stanislaw Halik | |
2016-08-16 | compat/options: unindent inside namespace only | Stanislaw Halik | |
2016-08-16 | compat/options: qDebug() in singleton dtor | Stanislaw Halik | |
This singleton is used around the codebase so make its lifetime clear to the developer. | |||
2016-08-16 | compat/options: cast to the right type for bigger than word things | Stanislaw Halik | |
2016-08-16 | compat/options: ensure all ini names sorted | Stanislaw Halik | |
2016-08-16 | compat/options: add option to reload all bundles | Stanislaw Halik | |
2016-08-15 | dinput, COM: temporary fixes, to be tested | Stanislaw Halik | |
2016-08-14 | compat/com: no need to initialize OLE at all times | Stanislaw Halik | |
2016-08-13 | default to apartment threading for now | Stanislaw Halik | |