Re: [gpsd-dev] Release this week (was: Re: ✘reset "Support UBX NAV-PVT" commit) 22a020ec1c2bc85eff681ecacc6d2bb79fdddc9c

[Fred Wright]

On Tue, 5 Sep 2017, Eric S. Raymond wrote:

> Gary E. Miller <address@hidden>:
> > Clark: can you resubmit in smaller pieces?
>
> I just received direction from ICEI to ship this week. Clark,
> whatever you want merged needs to get in soon if it's going to
> ship in this release.

One of the three Travis builds has been failing lately.  There had been a
couple of cases in the past where the failure was clearly due to a burp in
the testing infrastructure rather than in GPSD itself, so I didn't look
too closely at this failure when it first appeared, but it's been
consistent for about a month.

Although the failure first appeared when I pushed Clark's change, it's in
an area unaffected by the change, and remains after Gary's revert of that
change, so it's actually due to an external change rather than something
in GPSD, but it's a real build failure in certain configs.  It's
apparently due to an incompatibility between some versions of clang++ and
some versios of C++ headers, and only affects the C++ code, and has been
biting a number of people in other projects as well.  I've been unable to
reproduce it on any of my three Linux VMs, so testing a workaround would
be difficult.  Unless someone else can reproduce it, I may just put in a
hack to disable the C++ support in the Travis build that uses clang, in
order to get clean Travis results until such time as the error can be seen
in a more debuggable case.

Fred Wright