We want to be somewhat standards compliant if we can.
There is a small complication, which we work around:
gcc sets __STRICT_ANSI__ in "c++11" mode (the non-strict mode would be
"gnu++11"), which causes the "strdup" definition to go away. If wx < 3.1 is
compiled with GNU extensions, it just forwards the wxCRT strdup function to
system strdup and doesn't create a definition for the wxCRT function, and
when a program is then compiled in strict mode against that, the wxCRT
function is missing.
From 3.1 onwards, the function is provided unconditionally in wx, so the
workaround is no longer needed there.
This moves the generated files out of the source tree and into
the build directory. They are now regenerated each time they are
needed, based on the timestamp of the generated file compared to
the timestamp of the lemon file.
To do this, we also bundle lemon into the thirdparty directory
and build it for ourselves since it is a very tiny program and
not all platforms seem to distribute it in a consistent manner.
Fixes https://gitlab.com/kicad/code/kicad/issues/5013
This is required, because Ninja otherwise doesn't properly rebuild after
updating the version string header file.
Because BYPRODUCTS requires CMake 3.2, bump the minimum version and remove
now unused compatibility code for older CMake versions.
It is already added with MSVC, and must be added with gcc/mingw
However, the cmake option:
add_definitions( -DBOOST_UUID_FORCE_AUTO_LINK )
does not work on msys2 (at least with my cmake version), so another syntax is used.
* Split up the thirdparty code into the thirdparty folder (#3637)
* Create a new kimath static library containing all the math functions
This is part of cleaning the build system for #1906.
If an in-tree build was done on Linux, it would install
both the configured kicad.appdata.xml file and the template
kicad.appdata.xml.in file. This breaks packaging, and is not
needed.
CHANGED: Update the Linux appdata file to include new tags
* Include version tags (and others) in the appdata file
* Refactor the version string generation to clean it up
Remove not used files:
- markdown2html is not used (we are using "Sundown" tool)
- newstroke_font_without_CJK.cpp is removed.
Due to code optimization the full font works fine without memory issues
-fix also a few very minor compil warnings
It allow using the previous font set without CJK.
The new font (with CJK) is very large (10x), and can create (on Opengl) out of memory issues
with some graphic cards.
Our coroutine system can make debugging memory issues harder by not
following a standard stack allocation system. We can get around this by
using valgrind's built-in stack instrumentation. Each coroutine
registers a stack allocation allowing memcheck to recognize when
accesses are bounded.
KICAD_STDLIB_DEBUG option generate intrusive tests and asserts, and on Windows, generate crashed not captured by GDB.
KICAD_STDLIB_LIGHT_DEBUG generate less intrusive tests and asserts.
it adds only -Wp,-D_GLIBCXX_ASSERTIONS and that generate less asserts.
Add also the new build options in dialog about.
This adds std library assertions and address sanitizer options.
They are not yet enabled by default for debug builds as they are not
compatible with our inline assembler
Defines:
- inhibit generation of #pragma comment(lib, ...) from boost
- inhibit warnings about "unsafe" containers
- inhibit warnings about "unsafe" C functions
- inhibit warnings about "deprecated" POSIX functions
- suppress min/max macros from windows.h
Flags:
- source and execution charsets are UTF-8
- suppress warnings about throw() not being fully supported in the compiler
- suppress warnings about values being explicitly cast to bool
- enable string pooling
- enable unreferenced code removal
- enable COMDAT folding
- generate PDB debug information
Seth Hillbrand
Simon Richter
Mark Roszko
Ian McInerney
Jon Evans
Tomasz Wlostowski
jean-pierre charras
Wayne Stambaugh
Seth Hillbrand