because mingw-w64 is too old. The fix is done in the link I gave : https://sourceforge.net/p/mingw-w64/mailman/message/34033485/

no because gcc (for windows) adds it by default, otherwise you will have a bunch of functions which are not defined.

so if you don't care i'll abandon this revision and i'll do both fnmatch and regex removal, i'm fed up this takes so long

it's not my fault if you use an old mingw-w64 which does not have AddClipboardFormatListener (). this function is there since vista (cf https://docs.microsoft.com/sr-latn-rs/windows/desktop/api/winuser/nf-winuser-addclipboardformatlistener ) and has been in minw-w64 for 4 years (see https://sourceforge.net/p/mingw-w64/mailman/message/34033485/ ). It works perfectly on my computer with win7. So what is broken is not D8644, it's your CI for windows.

it will break if there is a CI for windows...

ok, you prefer another approach

i don't see anything in eldbus/meson.build about c++ files

otherwise, installation has finished without problem

what about meson ?

the fix should be backported to 1.22

$ git grep eldbus_eolian_cxx_hh
src/Makefile_Cxx.am:nodist_installed_eldbuscxxmainheaders_DATA = $(eldbus_eolian_cxx_hh) $(eldbus_eolian_cxx_impl_hh) lib/eldbus/Eldbus_Model.eo.hh

fixed with D8598

• remove unwanted comment

it's normal : D8543 must be applied first

see D8586

see D8586

that's my way of doing things : i add symbols only when needed.
if it is the way t go (adding a lib in eina so that all the libs/bins have that lib), then why not adding all the libs to eina and none to the other libs ?
there are several meson.build where some libs are explicitely added.

S you consider that, if byteswap.h exists, then the functions bswap_** exist ?

maybe actually this diff should be abandoned and i should remove the test in configure.ac about __builtin_lround ?

the autotools should also be updated (to verify the concordance of config.h), as well as eina_config.h.in, which then has useless macro

actually, __builtin_lround() is not used at all in the code ,afaics

can you be a bit more explicit, please ? i fear i don't understand what you want me to do

git grep HAVE_LROUND :

can't strncmp(pTemp, "br", 2) be replaced with 2 tests (same for "ps") ?

and we should open a task to mention we must have a test suite

imho, if cairo backend has bugs that the native one has not, and if it is not maintained anymore, it should just be dropped

btw, meson configure returns :

why is that option not necessary on linux ?

Fixed in D8432

btw, meson command :

What is problematic is that I want to compile the EFL with ewpi. and I don't know the computer on which a user can compile the EFL.

As I have said, using MSYS2 + mingw-w64 IS cross compilation : I have a windows 7 64 bits where I build the EFL for 32 bits. Then having cpu and cpu_family is important. I agree that endianness and system are not important.

you don't tell me how to set the properties for host_machine : system, cpu, cpu_family endiannes