Re: [gpsd-dev] Hang on Linux/Ubuntu

[Greg Troxel]

"Eric S. Raymond" <address@hidden> writes:

> Yes.  git pull and try again.  I just fixed a problem that caused a 
> repeatable 
> hang on Frank Nicholas's Mac OS X machine.
>
> Annoyingly, I still don't understand exactly why the problem was a
> problem.  The test frame used to write "# EOF\n" to gpsd when it
> detected that the last sentence of the test load had been sent, but
> for some unknown reason doing this did not unstick the select(2) call
> in gpsd's main loop under Mac OS X.
>
> I had a feeling that might cause some sort of timing issue and changed
> it to append "# EOF\n" to the test load when it's first built (if the
> -1 option is on)  That made the tests all work, and I think it likely
> to do the same thing for other *BSDs, but... 
>
> ...I don't know why.  There's some wacky interaction between select(2), 
> delays, and input timing that I don't grok, and it's variable by OS.

release-3.11-417-gc91368e passes cleanly on netbsd-6 i386 (once, which
is what I just tried).

What I was seeing was gpsfake spinning and gpsd idle.

I wonder if there was a missing fflush or something like that.

pgpkJRwZ_vkX0.pgp
Description: PGP signature