• Merge branch 'master' of https://git.enlightenment.org/core/efl
• Fix parsing of INF, NAN and a small value
• eina test: use fpclassify to test NAN and INF and do not test errno

it seems that whatever fix I do, i will never be able to implement a strtod so that it gives the same results than all the possible imlementations of strtod in the world (glibc, Windows libc, clang, newlib, freebsd, musl, maybe cygwin too, etc...).

about the first test which fails, it's not a problem of value, it's a problem of errno value which is not the same (no errno set contrary to strtod which sets errno to ERANGE)

on fedora, in the eina suite, only eina_convert fails.
I can comment the tests which fail if you want :-)

about the bugs, it's certainly something which can be fixed. but instead of having 0, this implementation return 2E-308, or the contrary (not a big deal imho, especially for a function which is never used).

after checking, with D7926, no problem with env functions, closing

can be abandoned after push of D7926 (D7926 also fixes meson)

I've check the compilation on Windows (MSYS2, mingw-w64, autotools) and Fedora. Compilation was fine on both OS. On Fedora, tests fail (when compared to strtod) on some rare values (around 2e-308) and for NAN

we need a strtod implementation. I have one that i have sent to @raster by mail.
this strtod function can be used for all platform

no problem on Windows side.

the fdo link in https://phab.enlightenment.org/w/arcanist/ is just not working at all

@ManMower if the question is for me, yes : Windows and Fedora

somme comments:

for now, i can't even compile the EFL on Windows, so...

fix wrong ]

• meson: remove wrong ]

i've tested the autotools before creating this diff

• meson: Improve description of 'windows-version' option

• Add windows-version option to meson build (like autotools)

@bu5hm4n you do not *find* what you target, you *set* what you target. It is the way to make available some features in the header files, at compile time, for the target we want. That's why there is no way to automatically know which version we target, especially if we cross compile on linux...

• Evil: simplify m4 code for the target Windows we want

On Windows (i7 6700K):

yes, i'vealso seen this in SO. I've changed the blist code directly...
now i'm recompiling the efl 1.21 with -O3 -march=native ...

on Windows :

someting like that :

we need to think of better ways

harfbuzz too

libwebp must be added

btw why linking in evil ?? the symbols are in eina...

msvcr100.a is provided by mingw64, but are we sure that the DLL is installed ? If not, the application will fail with function not defined. That can be a problem

you reverted the revert so you knew what you were doing.
otherwise, don't revert

Stefan helped me to fix issues with ewpi and ewpi compiles finely on fedora. So for me no problem on the ewpi side

i don't understand why using a function (using it only once) with a #ifdef instead of using the #ifdef directly...

i can't test anything until september, just look at code
i will abandon it and open another one, then, but in september

man strncpy says

there is no gl engines on Windows, so...

is there a unit test about this problem ?

sorry, i was on holydays and forget about your patch

getpwnam does not exist on Windows. it may be possible to get the home directory of a user. It is not straightforward, but i have seen some ideas about that

i plan to use C API for the ddraw engine, so there should be no problem in the future (around september/october)

you have not documented the behavior of the API function when the first argument is NULL or empty

that will certainly help, thank you

rename _eina_c_locale_get() to _eina_c_locale_new() and _eina_c_locale_free() to _eina_c_locale_del() to follow EFL conventions

no problem for me, the m4 diff is only a reordering of variables

add also several tests for exponential representation of the number

i've already implemented the clipboard support. Maybe change in the infra has removed it

I've just realized that you don't call freelocale() to deallocate the resources.

strtod_l does not exist on Windows, use _strtod_l instead

you didn't add the typedef

still error : line 457, the type is still locale_t

is there a test program so i can test this code ?

on Windows, newlocale() does not exist :

i will open a new ticket instead of trying to fix this mess, arcanist is too complicated for me

i used wrongly arcanist locally, so i created another patch, see D6478
closing this one

• Evil: fix usage of Evil in a program compiled with vc++
• Evil: fix usage of Evil in a program compiled with vc++ this should be enough to fix #T5206

@lauromoura but you said : "According to msys' pacman, I'm using mingw64/mingw-w64-x86_64-dbus 1.12.8-1." ?!?!?

i use d-bus from win-builds

patch update

• Merge branch 'master' of https://git.enlightenment.org/core/efl

ok, tests are run now, i have 12 successful tests and 15 which fail

• Tests: remove warnings when compiling tests if fork is not available and run tests properly on Windows

i would have done :

• Merge branch 'master' of https://git.enlightenment.org/core/efl

i've already broken the API on Windows several times. These functions are not used anymore and should not be used, even internally.

see D6363

hmm, arc diff fucked up this diff, i need help to restore the previous state...

• Evil: fix usage of Evil in a program compiled with vc++

• Merge branch 'master' of https://git.enlightenment.org/core/efl
• Merge branch 'master' of https://git.enlightenment.org/core/efl
• tests : remove warnings on Windows