On 12/1/08, James Carlson <james.d.carlson at sun.com> wrote: > I. Szczesniak writes: > > On 12/1/08, James Carlson <james.d.carlson at sun.com> wrote: > > > Roland Mainz writes: > > > > Is there a precedent that an application in OS/Net links with > > > > libsunmath.a ? > > > > > > > > > Only cmd/fps/fptest. > > > > > > > > > > The problem is that libm.so.2 lacks some functions (e.g. > > > > |iszero()|, |issubnormal()| etc.) and libsunmath.so.1 sits in > > > > /opt/SUNWspro/prod/lib/ ... ;-( > > > > > > > > > Note that depending on libsunmath.a likely means that OpenSolaris > > > participants can't easily compile with gcc. That seems like a bad > > > thing to me. > > > > I don't think so. gcc uses a builtin macro for iszero() and will not > > use libm or libsunmath. ON Makefiles can use compiler flags specific > > to the Sun compiler with cw -_cc=option. > > > It'll be ugly.
Do you know a better solution which can be implemented in a month or less? > That's why I said it was an architectural issue as well; if the > industry is going towards those functions (gcc or not), then we > shouldn't be hiding them away in some static library. > > Why not get the industry standard ones promoted to libm, and avoid the > hacks? How long will it take until the function is moved to libm? Irek
