rasmol @2.7.4.2: error: no such file or directory: 'DarwinMachineDefines'

[richard-evans (Dr. Richard Evans)]

Hi rasmol fails to build on M1/bigsur with the following error:

MOLDIR=\"/opt/local/share/rasmol/\" -DTHIRTYTWOBIT            -D_USE_RASMOL_H_     -c -o repres.o repres.c
:info:build clang: clang: clang: clangclang: : error: no such file or directory: 'DarwinMachineDefines'

It seems like an old definition that was recently removed in macOS, but I can't find any reference to it in the source code and google is unhelpful. Removing the definition "DarwinMachineDefines" in the makefile seems to fix it (it installs OK after that and runs fine) but not sure the best way to patch the port (bit of a port newbie) and whether removing this is in general safe.

Any thoughts?

Many thanks,

Richard

[ryandesign (Ryan Carsten Schmidt)]

Normally you should submit a main.log file when you encounter a build failure. But I tried a build on our Apple Silicon build machine and got the same problem, so that log is sufficient:

​https://build.macports.org/builders/ports-11_arm64-builder/builds/20536/steps/install-port/logs/stdio

Something is supposed to be defining DarwinMachineDefines to the set of flags that are appropriate for your system. That "something" was probably written before Apple Silicon machines existed, so it doesn't define DarwinMachineDefines to anything on your system, so the literal string DarwinMachineDefines gets inserted instead, which is wrong.

I haven't yet found what is (supposed to be) defining DarwinMachineDefines in this case. I did find it for example here:

​https://cgit.freedesktop.org/xorg/util/cf/tree/darwin.cf

And I found a previous reference to this kind of problem in #22528 (which was on Snow Leopard which was when Apple's gcc started defaulting to 64-bit compiles, so at that time it was probably the x86_64 architecture that wasn't being handled in the code.)

[ryandesign (Ryan Carsten Schmidt)]

Ok, here it is: it's xorg-cf-files:

$ port contents rdepof:rasmol | xargs grep DarwinMachineDefines 2>/dev/null
/opt/local/lib/X11/config/darwin.cf:# define DarwinMachineDefines   -D__powerpc__
/opt/local/lib/X11/config/darwin.cf:# define DarwinMachineDefines   -D__i386__
/opt/local/lib/X11/config/darwin.cf:# define DarwinMachineDefines   -D__x86_64__
/opt/local/lib/X11/config/darwin.cf:#define StandardDefines         DarwinMachineDefines -D__DARWIN__ \
$ port provides /opt/local/lib/X11/config/darwin.cf
/opt/local/lib/X11/config/darwin.cf is provided by: xorg-cf-files

Support for arm64 needs to be added to xorg-cf-files.

[ryandesign (Ryan Carsten Schmidt)]

Possibly as simple as this:

Imake.cf

@@ -37 +37 @@
 #  define x86_64DarwinArchitecture
 #  undef __x86_64__
 # endif
+# ifdef __arm64__
+#  define arm64DarwinArchitecture
+#  undef __arm64__
+# endif
 #endif
 
 #if defined(clipper) || defined(__clipper__)

darwin.cf

@@ -249 +249 @@
 # define ByteOrder              X_LITTLE_ENDIAN
 #endif /* x86_64DarwinArchitecture */
 
+#ifdef arm64DarwinArchitecture
+# define DarwinMachineDefines   -D__arm64__
+# if HasGcc3
+#  define OptimizedCDebugFlags  -Os
+# else
+#  define OptimizedCDebugFlags  -O2 -fno-strength-reduce
+# endif
+# define ByteOrder              X_LITTLE_ENDIAN
+#endif /* arm64DarwinArchitecture */
+
 #if OSMajorVersion < 7
 # define DarwinLocaleDefines    -DX_LOCALE
 #else

[richard-evans (Dr. Richard Evans)]

Sorry for continuing this ancient ticket! The problem still exists on Ventura, but I'm not sure how to regenerate the relevant file from xorg - is this run as part of the xorg installation? As I mentioned, removing the DarwinMachineDefines from the makefile fixes the bug for rasmol at least.