Keith Scroggins wrote:
So I built it (2.4.1) with GCC 2.95.3 and CVS MiNTLib. Running the tests, 5 failed:tprintf (no output, crashes with illegal instruction with "make check")td_sub (no output, just passed my first "make check" tried it again and it crashed with bus error)tset_ld (error with 5 or so lines of output) tcmp_ld (ERROR for NAN (1)) tget_sj (ERROR for get_uj + ERANGE + UINTMAX_MAX (2))I have not looked into any of these. My guess is Vincent's work with fixing the GMP issue will resolve tprintf.
Probably.
So, 5 out of 148 tests, thats pretty good I think.
Very good, since that libraries make a lot of math computations. I means that out system is quite reliable. The tests need to be fixed, though.
Remember, Mark said he had a lot of test failures with GCC 2.95, but only some failures with your GCC 4.x. Maybe that MPFR bugs would go away by using GCC 4.x ?
I'm looking into expanding some math functions in fdlibm, mainly to rectify some issues with building ScummVM, but maybe doing more after I get the initial functions I need installed.
Since there is no MiNT related stuf in any libm, we should definitely use the one provided by other active projects (glibc, Newlib...). The current versions probably have anything we need. If some useful functions are missing, we could add them and share it with the whole community. But doing non-MiNT related work just for us is a waste of time.
Also wanting to add some wchar support to MintLib for another item I have been working on porting for a LONG time. (can we say years?? :) ) Again, not something I am going to get done quickly.
Newlib has support for wchar, and it currently being improved. Again, there is nearly no MiNT specific code in wchar support.
-- Vincent Rivière