This is the mail archive of the
mailing list for the Cygwin project.
Re: cygwin 1.3.1 bug in fscanf("%lf", &a)
Christopher Faylor wrote:
> On Sat, Apr 28, 2001 at 12:46:18AM -0400, Charles S. Wilson wrote:
> >Christopher Faylor wrote:
> >> How about
> >> Subject: What gives with scanf in cygwin 1.31?
> >> That has a nice ring to it and I'm sure that the repetition will serve
> >> to motivate someone.
> >sarcasm aside, I don't think its necessary. I rebuilt a debuggable
> >kernel from current CVS (of both winsup and newlib) and the scanf()
> >functions work. Well, okay, I only explicitly tested vsscanf(), but
> >that should cover them all since __svfscanf_r() is used by all *scanf()
> >functions. I didn't single-step the new version in gdb, 'cause it
> >worked okay (why "debug" a working routine?)
> >The question is: how did this problem happen? I dunno -- chalk it up to
> >a thinko. I tested my changes -- but didn't explicitly test stuff that
> >I *knew* was already working -- like the %f scanning in the core routine
> >called by all *scanf() functions. Somehow, though, between my changes,
> >Jeff's changes, etc. %f parsing got broken. But it seems fixed now.
> Didn't Jeff's newlib changes just fix this?
Yes. That's what I meant when I said "it works now" using "current
CVS". Jeff was not SURE that the errors he corrected were the direct
cause of this problem; he said he'd try to look into it further this
weekend in case there was some deeper cause.
THEN my build of a debuggable kernel WITH his changes FINALLY finished,
and I ran my stupid test program. And it worked. I reported that
fact. Simultaneouly, Norman Vine reported that current CVS fixed some
problems folks were seeing in his port of ghostscript, and attributed
that success to Jeff's fixes in the *scanf() family.
I don't THINK further "debugging" or "motivation" is required. I don't
think further discussion on this thread is necessary. I think Jeff's
changes HAVE, in fact, licked this problem.
Want to unsubscribe from this list?
Check out: http://cygwin.com/ml/#unsubscribe-simple