V in Cygwin, "direct.h" no such file

[mrrstark]

Trying to compile on Cygwin (&MinGW) but getting this error:

...
Successfully compiled z-bitflag.c.
z-file.c:25:21: fatal error: direct.h: No such file or directory
compilation terminated.
Failed to compile z-file.c!

This is from rephial nightlies.
I followed these instructions in compiling.txt

Using Cygwin (with MinGW)
-------------------------

Use this option if you want to build a native Windows executable that
can run with or without Cygwin.

Use the Cygwin setup.exe to install the mingw-gcc-core package and any
dependencies suggested by the installer.

Run these commands:

./autogen.sh
./configure --enable-win --with-no-install --host=i686-pc-mingw32
make

[FunkyGibbon]

Lurker turned Poster

I've looked at the source, turns out whoever is committing that file is using (most likely) DMC to compile (digital mars compiler) only that set has this 'direct.h' file and it is NOT POSIX or FOS compliant, it will not compile on any GCC compiler on Linux, Windows or OSX.

Strip that include 'direct.h' out and it should compile fine and without errors.

A note: Why is a non-FOS compiler being used in an Open Source project? DMC is only freeware and only on Windows.

[takkaria]

Lurker turned Poster

I've looked at the source, turns out whoever is committing that file is using (most likely) DMC to compile (digital mars compiler) only that set has this 'direct.h' file and it is NOT POSIX or FOS compliant, it will not compile on any GCC compiler on Linux, Windows or OSX.

Strip that include 'direct.h' out and it should compile fine and without errors.

A note: Why is a non-FOS compiler being used in an Open Source project? DMC is only freeware and only on Windows.

I don't think anyone is using DMC. That line works fine with mingw - it's a standard Windows hedaer - just not on cygwin.

[FunkyGibbon]

Ah! I wasn't aware it was also a windows header, I've never seen it not even on windows.

Perhaps a simple ifdef would be a good choice? Because to compile that on *nix you would have to strip out that include. And it would avoid these compilation issues with people on other OS's.

#include (standard cross platform headers)

#ifdef _WINDOWS
#include direct.h (and other includes needed)
#else
#include (same includes but won't include direct.h)
#endif

[takkaria]

Ah! I wasn't aware it was also a windows header, I've never seen it not even on windows.

Perhaps a simple ifdef would be a good choice? Because to compile that on *nix you would have to strip out that include. And it would avoid these compilation issues with people on other OS's.

#include (standard cross platform headers)

#ifdef _WINDOWS
#include direct.h (and other includes needed)
#else
#include (same includes but won't include direct.h)
#endif

If you look in the file, this is already there.

[FunkyGibbon]

Well I see the problem, there are only ifdefs for Windows, none for Linux and none for Cygwin and none for OS X, so all it is doing is checking if it is Windows, if it is not there are no ifdefs to prevent any other OS from not including direct.h. Maybe something in the makefile also? (I like to use IDE's not a makefile).

#ifdef WINDOWS
# include <windows.h>
# include <io.h>
# include <direct.h>
#elif __GNU_LINUX__
# include (all includes without windows specific)
#elif __CYGWIN__
# include (all includes with windows specific headers but without direct.h)
#elif __APPLE__ && __MACH__
# include (all includes without windows specific, probably same as Linux includes)
#endif

That would probably solve the issue, though I am no where near my Mac to try it. I am quite used to using single batch ifdefs, especially when compiling code written by mostly Windows using or MinGW using developers. I usually do it this way and I don't have problems.

The main issue is that the direct.h handles the mkdir and so *nix cannot 'mkdir' unless it hits this code below the ifdef. It seems to just fail when hitting the direct.h and it does not look like the compiler gives 2 hoots about the windows ifdef.

Just my 2 cents

[takkaria]

Well I see the problem, there are only ifdefs for Windows, none for Linux and none for Cygwin and none for OS X, so all it is doing is checking if it is Windows, if it is not there are no ifdefs to prevent any other OS from not including direct.h. Maybe something in the makefile also? (I like to use IDE's not a makefile).

#ifdef WINDOWS
# include <windows.h>
# include <io.h>
# include <direct.h>
#elif __GNU_LINUX__
# include (all includes without windows specific)
#elif __CYGWIN__
# include (all includes with windows specific headers but without direct.h)
#elif __APPLE__ && __MACH__
# include (all includes without windows specific, probably same as Linux includes)
#endif

That would probably solve the issue, though I am no where near my Mac to try it. I am quite used to using single batch ifdefs, especially when compiling code written by mostly Windows using or MinGW using developers. I usually do it this way and I don't have problems

Just my 2 cents

If you look a bit more closely, you'll see that there are HAVE_*_H ifdef chunks around the relevant header includes. These are set up by either h-basic.h or autoconf or the makefile depending on the platform. The things we want to include are not based on platform but on header file availability, which is what the current setup does.

Obviously we need to stop direct.h being included if we're in cygwin and make sure that HAVE_DIRENT_H is defined in the makefile instead. But we already do what you're suggesting, we just need to tweak it.

[mrrstark]

So for cygwin, can I just comment out the direct.h, or define something to finish the compilation?


I also tried a MinGW shell compile, following the instructions in compiling.txt, but I got some other error, and no files compiled at all. I'm not at home atm, so can't provide the error message here. Is it expected to work as in compiling.txt?

[Nick]

I also tried a MinGW shell compile, following the instructions in compiling.txt, but I got some other error, and no files compiled at all. I'm not at home atm, so can't provide the error message here. Is it expected to work as in compiling.txt?

In MinGW, I just use

MINGW=yes make -f Makefile.win

from the src directory.

[mrrstark]

In MinGW, I just use

MINGW=yes make -f Makefile.win

from the src directory.

Awesome. That worked.

[MattB]

I'm not really sure why, but I always read every line of these posts, and I genuinely haven't understood a single word so far! Must be the Angband effect, I guess...

[AnonymousHero]

I'm not really sure why, but I always read every line of these posts, and I genuinely haven't understood a single word so far! Must be the Angband effect, I guess...

I think it's the Lack of Arcane Knowledge That Should Be Mostly Useless for Anyone Developing Software By Now, but Regretfully Isn't. I call it the LoAKTSBMUfADSBNbRI principle, but unfortunately that term hasn't caught on.

... but take heart and don't fret! Soon such "knowledge" won't be required!