Open TTD can't be compiled with MinGW

[Joe#]

Hi,

Seeing a wiki tutorial page for doing this, I had a go at compiling OpenTTD for my Windows 8.1 64bit OS yesterday but I couldn't because MinGW's implementation of the POSIX header file "unistd.h" does not implement: fdatasync(int) which is called by "ini.cpp" at line 84.

To makesure there wasn't something I was doing wrong I sent an email to the MinGW users mailing list and Gisle replied:

fsync() nor fdatasync() isn't part of MingW, but can easily
be replaced by:

#include <io.h>

extern __inline__ int fdatasync (int fd)
{
return _commit(fd);
}

(or a macro).

The POSIX distinction between fsync() and fdatasync() is probably
not important for OpenTTD. Looking at:
http://git.openttd.org/?p=trunk.git;a=b ... da;hb=HEAD

the code around 'WITH_FDATASYNC' looks fishy.

Ref:
https://msdn.microsoft.com/en-us/librar ... 2147217396

So I think the Wiki page for the MinGW tutorial should be disabled until MinGW can compile OpenTTD.

[Joe#]

Kieth from the MinGW users mailing list:

> fsync() nor fdatasync() isn't part of MingW,

They aren't even guaranteed to be supported by a POSIX conforming
system, since both are designated as "optional", (which means that
applications wishing to use them should check the applicable feature
test macros, to ensure that they are provided); since MinGW defines
*neither* _POSIX_FSYNC *nor* _POSIX_SYNCHRONIZED_IO, it should be clear
that neither function will be supported, and there is therefore no
requirement, (even under POSIX.1-2008), to provide so much as even a
stub implementation.

> ...snip...
>
> the code around 'WITH_FDATASYNC' looks fishy.

Agreed; very fishy indeed, for there appears to be no attempt whatsoever
to check API availability, before gratuitously assuming support.

[Eddi]

you should probably post that on the bug tracker, so a dev can review it.