Hello,
as stated in the title, I’m trying to cross compile ruby 2.1.0 on my
Linux machine (Fedora 19) for Windows using the mingw32-w64 toolchain.
My current configure line looks as follows:
./configure --prefix=$MY_PREF --host=i686-w64-mingw32
–disable-dtrace --with-baseruby=/usr/bin/ruby-mri
I needed to disable dtrace because it caused a non-existant header
(sys/sdt.h) to be included.
At first, everything seems to go fine, the core .c files are compiled,
but then the Makefile builds the windows executable “miniruby.exe” and
continues to attempt generating more file with it. Running an exe causes
wine to automatically be invoked for it, which is pretty crazy.
Unfortunately, not even that works, as miniruby.exe always seems to get
stuck on a mutex no matter what:
err:ntdll:RtlpWaitForCriticalSection section 0x5380e90 "?" wait
timed out in thread 0009, blocked by 0000, retrying (60 sec)
That being said, it should have never tried to run a windows executable
in the first place I think. As far as I understand, ruby needs another
ruby to build itself, and to remedy the situation where no ruby exists
on the building system, it builds a minimal ruby to progress. However,
this makes no sense when cross compiling because any generated
executable will be non-native for your build platform.
As you can see, I tried supplying the configure script with a “base
ruby” path, hoping it would use that instead of creating miniruby.exe,
but the option didn’t change anything. I also tried manually editing the
generated Makefile and chaning the MINIRUBY variable declaration to
point to ‘/usr/bin/ruby-mri’ instead, but then I encountered a different
error:
generating msvcrt-ruby210.def
./win32/mkexports.rb:19:in `create': unsupported platform:
x86_64-linux (ArgumentError)
I’m assuming this is because the script in question somehow extracted
the platform out of the ruby executable.
I’m thoroughly lost at this point. I also tried commenting out all
EnterCriticalSection and LeaveCriticalSection calls in win32/win32.c,
compiling miniruby.exe with it, and changing win32.c back, but I soon
hit the next error:
./win32/mkexports.rb:147:in foreach': No such file or directory - i686-w64-mingw32-nm --extern --defined libmsvcrt-ruby210-static.a (Errno::ENOENT) from ./win32/mkexports.rb:147:ineach_line’
from ./win32/mkexports.rb:89:in objdump' from ./win32/mkexports.rb:154:ineach_export’
from ./win32/mkexports.rb:163:in each_export' from ./win32/mkexports.rb:40:ininitialize’
from ./win32/mkexports.rb:21:in new' from ./win32/mkexports.rb:21:increate’
from ./win32/mkexports.rb:25:in extract' from ./win32/mkexports.rb:170:inblock in ’
This might very well be caused by the mutex locks I commented out. This
is as far as I’ve gotten.