The autogen.sh script was made executable; it also now does not annoy…#2
The autogen.sh script was made executable; it also now does not annoy…#2OhGodAPet wants to merge 1 commit intodecred:3.7from OhGodAPet:3.7
Conversation
…ing configure for you (and do it wrong in the process.) The sleep() function was removed from compat.h - I use a custom MinGW-w64 toolchain with pthreads-w32, and sleep() is provided just fine. If left there, it will cause linker errors for the multiply defined symbol. The configure.ac file (which generates the configure script) now includes an option to disable the previously mandatory ADL header file checks - this check will error out unconditionally when cross compiling. It may now be disabled by --disable-adl-checks, which defaults to off (meaning the checks stay on by default.) The configure script also now will not assume winpthread is the ONLY kind of pthread library you can use on Windows, and will check for both winpthreads and a REAL Windows pthreads library (like pthreads-w32, which I use because it supplies more of the pthread library than winpthreads) - it will then handle linking accordingly, i.e. -lwinpthread in the former case, and -lpthread in the latter. The variable has_winpthread is still set to true if another Windows pthread lib is found - this is not a mistake, it is required for the code to work properly. Someone should rename that at some point.
|
So this PR got an OK from @jolan (in that it is okay with the linux builds). Just going to get one of us to test it on windows then we should be good to go. Sorry for not getting to it yesterday. |
|
Sorry for what? I'm also hosting the patch (signed by me) on my site, and posted this on the forums - anyone who wants to use it can. Take your time, ensure there's no breakage. |
|
@wolf9466 I'm pretty sure PR#3 takes care of everything this one takes care of as well (and a few other things so I merged it in). Sorry I didn't get yours in but thanks for getting the ball rolling on fixing up the config stuff! |
…ing configure for you (and do it wrong in the process.) The sleep() function was removed from compat.h - I use a custom MinGW-w64 toolchain with pthreads-w32, and sleep() is provided just fine. If left there, it will cause linker errors for the multiply defined symbol. The configure.ac file (which generates the configure script) now includes an option to disable the previously mandatory ADL header file checks - this check will error out unconditionally when cross compiling. It may now be disabled by --disable-adl-checks, which defaults to off (meaning the checks stay on by default.) The configure script also now will not assume winpthread is the ONLY kind of pthread library you can use on Windows, and will check for both winpthreads and a REAL Windows pthreads library (like pthreads-w32, which I use because it supplies more of the pthread library than winpthreads) - it will then handle linking accordingly, i.e. -lwinpthread in the former case, and -lpthread in the latter. The variable has_winpthread is still set to true if another Windows pthread lib is found - this is not a mistake, it is required for the code to work properly. Someone should rename that at some point.