The Sfio And The Bizarre

pudge on 2002-05-10T12:21:00

So Matthias asked me, well, if we don't need SFIO for MacPerl's IO anymore, can we just not link with it it? I don't know, I replied, but ... doesn't SFIO provide our implementation of sprintf? I thought perl supplied sprintf now, he said. Sorta, I responded, but not for %e, %f, %g. Those are still handled with your local implementation.

I don't know if not linking to SFIO would cause other problems -- I'd probably need to fix a few things here and there, plus the MacPerl app still relies on it for one or two things -- but we need to link to it if for no other reason that we need it for sprintf. Yow!


efg egf feg fge gef gfe

jhi on 2002-05-10T22:04:19

Sooo... are you saying that

  1. SFIO really does come with an (IEEE-float-aware) bit-level implementation of floating point output?
  2. There is no %efg support at all in the MacOS Classic "libc"? (whatever that means in MacOS Classicese...)
  3. both?

Re:efg egf feg fge gef gfe

pudge on 2002-05-10T23:23:29

SFIO does implement efg, though not very well (there are a few bugs, at least in the SFIO version we use, which is 98 IIRC). Mac OS's standard library has no efg support, or so I am told.

Re:efg egf feg fge gef gfe

jhi on 2002-05-11T20:50:42

That MacOS standard library does no efg is a bit hard to believe. But having never programmed a single C program in MacOS Classic, who am I to argue :-)

Re:efg egf feg fge gef gfe

pudge on 2002-05-11T21:21:18

Yeah, I found it a bit odd too. However, if I ever feel inclined to try to completely sever the tie to sfio, I'll look more into it. We have PerlIO working now, which is the real objective.