4.4BSD/usr/share/man/cat3f/trapov.0

Compare this file to the similar file:
Show the results in this format:

F>=1  .if 3F<=8 .    if 3F>1 .if 3F<6 .ds an-extra3 "BSD Program-



TRAPOV(3F)             BSD Reference Manual            TRAPOV(3F)

mer's Manual
NNAAMMEE
       trapov - trap and repair floating point overflow

SSYYNNOOPPSSIISS
       ssuubbrroouuttiinnee ttrraappoovv ((nnuummeessgg,, rrttnnvvaall))
       ddoouubbllee pprreecciissiioonn rrttnnvvaall

DDEESSCCRRIIPPTTIIOONN
       NNOOTTEE:: TThhiiss rroouuttiinnee aapppplliieess oonnllyy ttoo tthhee oollddeerr VVAAXX 1111//778800''ss..
       VVAAXX  ccoommppuutteerrss  mmaaddee  oorr uuppggrraaddeedd ssiinnccee sspprriinngg 11998833 hhaannddllee
       eerrrroorrss ddiiffffeerreennttllyy..  See _t_r_p_f_p_e(3F) for  the  newer  error
       handler.   This routine has always been ineffective on the
       VAX 11/750.  It is a null routine on the PDP11.

       This call sets  up  signal  handlers  to  trap  arithmetic
       exceptions  and  the  use  of  illegal operands.  Trapping
       arithmetic exceptions allows the user's program to proceed
       from  instances  of  floating  point overflow or divide by
       zero.  The result of such operations will  be  an  illegal
       floating  point  value.  The subsequent use of the illegal
       operand will be trapped and the operand  replaced  by  the
       specified value.

       The  first  _n_u_m_e_s_g  occurrences of a floating point arith-
       metic error will cause a message  to  be  written  to  the
       standard  error file.  If the resulting value is used, the
       value given for _r_t_n_v_a_l will replace  the  illegal  operand
       generated  by the arithmetic error.  _R_t_n_v_a_l must be a dou-
       ble precision value. For example, ``0d0'' or ``dflmax()''.

FFIILLEESS
       /usr/lib/libF77.a

SSEEEE AALLSSOO
       trpfpe(3F), signal(3F), range(3F)

BBUUGGSS
       Other   arithmetic  exceptions  can  be  trapped  but  not
       repaired.

       There is no way to distinguish between an integer value of
       32768 and the illegal floating point form.  Therefore such
       an integer value may get replaced while repairing the  use
       of an illegal operand.









4.2 Berkeley Distribution  June 5, 1993                         1