mirrors/ecl
1
Fork 0
mirror of https://gitlab.com/embeddable-common-lisp/ecl.git synced 2026-01-14 05:12:38 -08:00
Code Issues Projects Releases Packages Wiki Activity

Final fix for a problem with buggy tanf().

Browse source
This commit is contained in:
jjgarcia 2008-08-03 14:54:56 +00:00
parent 1743a8a68a
commit 2eae8071fa
1 changed files with 8 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

16
src/c/num_sfun.d
View file

@ -600,13 +600,13 @@ cl_cos(cl_object x)
/*
* As of 2006-10-13 I found this bug in GLIBC's tanf, which overflows
* when the argument is pi/4. It is 2008 and this has not yet been
* solved.
* solved. Not only that, but if we use tan() on float, GCC automatically
* and stupidly forces the use of tanf().
*/
#if defined(__amd64__) && defined(__GLIBC__)
# ifdef tanf
# undef tanf
# endif
# define tanf(x) (float)tan(x)
static double safe_tanf(double x) { return tan(x); }
#else
# define safe_tanf(x) tanf(x)
#endif
cl_object
@ -618,13 +618,13 @@ cl_tan(cl_object x)
case t_fixnum:
case t_bignum:
case t_ratio:
output = ecl_make_singlefloat(tanf(number_to_float(x))); break;
output = ecl_make_singlefloat(safe_tanf(number_to_float(x))); break;
#ifdef ECL_SHORT_FLOAT
case t_shortfloat:
output = make_shortfloat(tanf(ecl_short_float(x))); break;
output = make_shortfloat(safe_tanf(ecl_short_float(x))); break;
#endif
case t_singlefloat:
output = ecl_make_singlefloat(tanf(sf(x))); break;
output = ecl_make_singlefloat(safe_tanf(sf(x))); break;
case t_doublefloat:
output = ecl_make_doublefloat(tan(df(x))); break;
#ifdef ECL_LONG_FLOAT