Re: ✘gpsd .23.2~rc1

[Gary E. Miller]

Yo Hal!

On Sat, 09 Apr 2022 13:27:48 -0700
Hal Murray <halmurray@sonic.net> wrote:

> >> This happens on 32 bit Debian and 32 bit NetBSD.  
> > But still a libc error.  printf() is supposed to round, and it is
> > not doing so.  I would not like to change all printf() uses of %f.  
> 
> You are claiming it is a rounding error in printf.  I tried to make a
> test case.  I can't get it to fail.  I'm down to the last bit.

Fred claims it may be related to some compilers promoting double to
long double, and then things round differetnly (and not in compliance
with IEEE 754).  Since he ahs ideas, please discuss it with him.

> Any suggestions on how to get more info?

My universal debug suggestion is always: more printf().

> I think I want do do something like change the printout to print in
> hex and make a chk file on a good system then compare with the system
> that fails.

I'm sure the hex will also differ.  If Fred is right, the rounding
error happened earlier, in an unexpected long double to doubble
conversion.. Which would be a violation of C standards, but it happens.

> How do I make a .chk file?

$ scons check
$ cd buildtmpe
$ ./regress-driver -b test/daemon/XXX.log





RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
        gem@rellim.com  Tel:+1 541 382 8588

            Veritas liberabit vos. -- Quid est veritas?
    "If you can't measure it, you can't improve it." - Lord Kelvin

pgp1QNEG8BmM3.pgp
Description: OpenPGP digital signature