Re: [gpsd-dev] regression tests blow up on os x 10.9 with shm issues

[Eric S. Raymond]

Greg Troxel <address@hidden>:
> Things are much better with release-3.11-434-g8b107ca, passing on os x
> 10.9, netbsd-5 amd64 and netbsd-6 i386.

That's great!  Have you played with WRITE_PAD to see if it can be zeroed or
reduced?

Interestingly, when I try building on Frank Nicholas's OS X box, GPS regression
tests pass but gpsdecode coredumps when trying to interpret RTCM2 or AIVDM
files.  I suspect this is some kind of dynamic-loading issue; I prevented it
once by forcing static linking, but have not been able to reproduce that.

> I am seeing one leftover shm segment after running tests.  Not a huge
> problem, but really there should not be any.

I can think of a good reason for one to be left over. Apparently shmctl(2)
marks segments for deletion but it's up to the kernel to decide when the
memory is actually garbage-collected - that's what the man page description
implies, anyway.

The only known issue left is Hal's bizarre build glitch.  

I might try tweaking the build to work with uClibc - libdbus needs to move
to the end of the library list - but it looks to me like the C code is ready
to ship.  I'll run all four static checkers on it, though.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>

signature.asc
Description: Digital signature