Support for w*ndows

[Torbjorn Granlund]

nisse at lysator.liu.se (Niels Möller) writes:

  Torbjorn Granlund <tg at gmplib.org> writes:
  
  > I put on my TODO list after you sent your initial message to setup
  > nightly testing using wine.  Does user-level instructions run natively
  > on the CPU, or is there qemu-like translation going on?
  
  I'm fairly sure the instructions run natively. *Maybe* you can hook in
  some cpu emulator if the native system is not of the same type. Still,
  there seems to be a lot of overhead, running make check under wine takes
  5:10 here, to be compared to 1:38 for an elf 32-bit build on the same
  hardware. I'm not familiar with how wine really works, so I can't say
  where this difference comes from, or if it should be expected.
  
Perhaps use TESTS_ENVIRONMENT=time and TESTS_ENVIRONMENT="time wine" to
see if some particular test run slowly?  I expect I/O to needs
interpretation, while pure computations should run fast.

  > You are of course most welcome to debug these failures...
  
  I wonder if one can build a wine-targeted gdb...
  
I debugged my recent Windos work using printf statements and asserts...
It slows down debugging by perhaps 10x, but it works.

  And it would be nice if someone who really cares about windows support
  could test on a native system to see if it fails in the same way.
  
But that would be non-conclusive, unless you have the same libs,
compiler, assembler, linker...

-- 
Torbjörn