[gnu.gcc.bug] gcc 1.36 always uses software floating-point math libraries

ghfeil@white.toronto.edu (Georg Feil) (11/14/89)

On a sun3 with OS 4.0.1, gcc 1.36 always links the default libm.a if you
specify -lm even when the 68881 or FPA are being used. To fix this, the 
definition of LIB_SPEC in config/tm-sun3.h should include a -L option
as follows:

    for -msoft-float,   add  -L/usr/lib/fsoft  to LIB_SPEC
    for -m68881     ,   add  -L/usr/lib/f68881 to LIB_SPEC
    for -mfpa       ,   add  -L/usr/lib/ffpa   to LIB_SPEC

Otherwise users must remember to enter the correct -L specification
each time they compile a program that uses the math lib.

Georg.
-- 
Georg Feil                                            "The roof is leaking
ghfeil@white.toronto.edu                               and the wind is howling"
...if that doesn't work try one of:
{uunet,pyramid,watmath,utzoo}!utcsri!white!ghfeil
ghfeil%white.toronto.edu@relay.cs.net     (ARPA)