cc: Kevin
In message <[email protected]> "Ralf W. Grosse-Kunstleve"
The reason why I ask is that 64-bit Coot fails to read MTZ files [1]
That's weird. We are using the cmtz library essentially directly from the current ccp4 cvs. Reading and writing mtz files is routinely and heavily exercised on all platforms we support, and is known to work flawlessly everywhere, including a bunch of 64-bit platforms. It is not too long ago that I ran the tests with valgrind, which reported 0 errors.
Are you compiling with -fno-strict-aliasing? That could in theory be important.
If that doesn't help, what are the symptoms of the mtz read failure?
Hi Ralf, thanks for getting back to me. This info is mostly second-hand of course, I don't have a usable 64 bit machine to hand. These are the symptoms so far (as far as I can gather). It fails (hangs) in fftw (IIRC) and libm (different cases) 0x0000003486b13885 in __ieee754_pow () from /lib64/tls/libm.so.6 (gdb) where #0 0x0000003486b13885 in __ieee754_pow () from /lib64/tls/libm.so.6 #1 0x0000003486b144b8 in __ieee754_pow () from /lib64/tls/libm.so.6 #2 0x0000003486b27903 in pow () from /lib64/tls/libm.so.6 #3 0x0000002a9641b18f in clipper::Grid_sampling::init (this=0x7fbfffc190, spacegroup=Variable "spacegroup" is not available. ) at coords.cpp:399 Yes, indeed I do compile clipper with -fno-strict-aliasing, for both the GTK1 and GTK2 versions. (not anything else though). I did hear that 64 bit GTK1 version works with MTZ files (or can be made to work) but GTK2 version does not (above is gtk2 version). I did try to track this down with Kevin a while ago, the clipper test program for fft works fine with the test data in 64 bits. But not in Coot. Then my disk died and I never got back to booting 64 bit OS. Cheers, Paul.