[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gpsd-dev] Preparing for release
From: |
Eric S. Raymond |
Subject: |
Re: [gpsd-dev] Preparing for release |
Date: |
Wed, 18 Feb 2015 13:51:26 -0500 |
User-agent: |
Mutt/1.5.23 (2014-03-12) |
Greg Troxel <address@hidden>:
> > Does anyone running on a Mac have any theory or insight about what's
> > going on? Greg Troxel's reports seem to indicate that the first open
> > of Mac tty devices is doing something to them that is making later
> > attempts to use them fail, but also that it is not the recent change to
> > blocking I/O. I've been hoping we could see before-and after stty
> > reports, but nobody has generated any.
>
> I don't think it's a first open issue. I tested also with a ublox 4
> eval board (old, from 2008), and it seems to basically work, even
> stopping/starting gpsd. Something seems funky about ublox6 in
> particular.
While this is theoretically possible, I have two uBlox-6 based devices
I test with regularly (one GR-601W and one uBlox EVK-6H eval kit) and
they're as solid and reliable as granite with my Linux systems - a real
pleasure to work with. I think the EVK-6H is my favorite piece of test
equipment ever.
We're also not getting this weirdness from *BSD systems other than OS
X, AFAIK, So if there's something funky about uBlox-6, it seems to be
specifically in interaction with the Mac.
> I now think that the sometimes-works-initially behavior is not about
> serial port reset issues but about catching the device at the right time
> in power up sequence.
By "the device", do you mean the GPS or the host-side serial hardware?
Because if you mean the GPS, then the salient question is why Linux
never hits this timing window.
> I also don't think it's necessary or reasonable to declare the Mac as
> "unsupported".
Again, if there's something funky about uBlox-6 it's specifically in
interaction with the Mac. And we have Frank Nicholas's report of
continuing Mac issues with a Bluetooth-based GPS, and we don't know
that the Mac PL2023 bug of 2009 was ever fixed.
It's not for one of these that I just took Mac OS X off the list,
it's for all three of them. I'm open to putting it back on if,
for example, we discover that there's some Mac-specific invocation
we can mutter that fixes the first two problems.
> It's really just that there is a bug with at least a
> device or two and we don't know how bad the issues are, and it may be
> that the bug is in mac code instead of gpsd. So just saying that seems
> far more useful. I will try a few more devices.
If we think there are substantial odds of a Mac OS bug rather than a
GPSD problem, what alternative do we have to declaring non-support
until we know what's going on? Shipping code that we know will be
flaky *and calling the platform supported* doesn't sit well with me.
I feel a need to be conservative about this. People rely on GPSD for
life-critical uses; when we say a platform is "supported" I don't want
there to be any doubt or fuzziness about what that means.
--
<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>
signature.asc
Description: Digital signature