gpsd-dev
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: ✘gpsd .23.2~rc1


From: Gary E. Miller
Subject: Re: ✘gpsd .23.2~rc1
Date: Wed, 13 Apr 2022 10:43:56 -0700

Yo Hal!

On Wed, 13 Apr 2022 10:34:49 -0700
Hal Murray <halmurray@sonic.net> wrote:

> > Thus compile options will not affect the pre-compiled library...   
> 
> How much floating point work does GPSD do in library routines?

hypot(), sin, cos(), modf(), etc.  Worse a lot of what you think
the C compiler does, is actually library calls.

When the receiver provides all the data, very little, but when the
receeiver only sends a little data, gpsd tries to compute the
missing parts.  Like, in this case, speed.

> That case worked before you added the precision flags.
> It was a different case that failed.

"scons check" stops at the first failure.  So how do you know the later
failure was working?  UNless you reverted the change and directlyran
that test.

> > Pointing again to a musl issue.  
> 
> You are confusing things.

Easy, if I wasn't confused, it would be fixed.  Not enough coffee
yet this morning.

>  The musl problem was with strerror_r.

And that is fixed.  Are the regressions good on musl now?

> This floating round problem shows up on a few 32 bit systems: Debian
> and NetBSD.

So both gcc and clang?

What about OpenBSD?

Do those all use FLT_EVAL_METHOD == 2?  Which means non-standard
floating point math.

> I haven't seen it on Raspberry Pi.

Ditto.

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

Attachment: pgpTFONiHGWht.pgp
Description: OpenPGP digital signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]