Support DragonFly BSD too?

[peternewman]

@RenZ0 as our chief *BSDer, should we support DragonFly BSD too?
https://en.wikipedia.org/wiki/DragonFly_BSD

[RenZ0]

I love the idea, I will have a look asap !

[RenZ0]

Here we go ! For a first try I replace the ifdef FreeBSD by DragonFly but we have a problem ;

/bin/sh ./libtool  --tag=CXX    --mode=link g++ -I./include  -I./include  -Wall -Wformat -W -fvisibility-inlines-hidden  -I/usr/local/include -pthread   -g -O2 -std=gnu++11 -pthread  -L/usr/local/lib/ -o examples/ola_dev_info examples/ola-client.o common/libolacommon.la  ola/libola.la 
libtool: link: g++ -I./include -I./include -Wall -Wformat -W -fvisibility-inlines-hidden -I/usr/local/include -pthread -g -O2 -std=gnu++11 -pthread -o examples/.libs/ola_dev_info examples/ola-client.o  -L/usr/local/lib/ common/.libs/libolacommon.so -L/usr/local/lib ola/.libs/libola.so /home/renzo/ola/common/.libs/libolacommon.so -lprotobuf -lpthread -pthread -Wl,-rpath -Wl,/usr/local/lib
common/.libs/libolacommon.so: error: undefined reference to 'backtrace'
common/.libs/libolacommon.so: error: undefined reference to 'backtrace_symbols_fd'
collect2: error: ld returned 1 exit status
*** Error code 1

in OLA code we have
$ grep -r 'backtrace' *
common/base/Init.cpp:  size_t size = backtrace(array, STACK_SIZE);
common/base/Init.cpp:  backtrace_symbols_fd(array, size, STDERR_FILENO);
configure.ac:# FreeBSD required -lexecinfo to call backtrace - checking for presence of
configure.ac:AC_CHECK_LIB([execinfo], [backtrace], [use_lexecinfo="yes"])

$ uname -irs
DragonFly 5.2-RELEASE X86_64_GENERIC
$ 
$ ls /usr/lib/libexec*
/usr/lib/libexecinfo.a		/usr/lib/libexecinfo.so		/usr/lib/libexecinfo.so.1

[peternewman]

We need config.log too. DragonFly has backtrace from the looks of things:
https://www.dragonflybsd.org/cgi/web-man?command=backtrace&section=3

Via libexecinfo, so it looks like you need to extend the case statement at line 390 of configure.ac too.

[peternewman]

You'll need to do this for the ifdefs to make them or:
https://stackoverflow.com/a/9682617

[RenZ0]

After some more fighting, it's working !
The configure.ac did it for backtrace.
Of course I add to apply the ossp-uuid patch, as it's not synchronized yet with FreeBSD.

Weird, make didn't find curses.h so I fixed with
ln -s /usr/local/include/ncurses/curses.h /usr/local/include/

Good news, DragonFly BSD has no problem with reinterpret_cast NULL.
There is some network checks fails too, even if root, see attached.

config.log
test-suite.log
root_test-suite.log

After a few tests,
olad seems ok, with web ui, ola_dmxmonitor too (to check curses), and python wrapper works too.

[peternewman]

@RenZ0 That's good news. Does DragonFly sync up eventually then, so the fix will appear given time?

We should fix the curses.h stuff properly too, it looks like it just needs some AC_CHECK_HEADER and then some #ifdef to include the correct header file.

Starting with the easiest one, the FQDN failure, what does hostname -f show on the command line?

I note from the network test log, we need to fix up the code so it fetches the sdl_type and sdl_index on DragonFly and possibly FreeBSD too, as there is a small loss of functionality there.

I'm not immediately sure about the TCPConnectorTester failure. I suspect the thread stuff is the same as FreeBSD (or similar), so we should fix that first. It's certainly throwing another operation not permitted error, and it works when run as root.

Can we have a pull request again please?

[RenZ0]

I am not sure when DragonFly sync up, I will ask to the team.
About FQDN, hostname was blank, I set it and now the NetworkTester is ok. For info, -f is not a valid option for bsd.
About curses.h, may I have some help ?
Sure I will do a PR, I still need to see how to include both freebsd and dragonfly in the same case statement in configure.ac ; easier to add a second case ?

[peternewman]

Okay.

Interesting that it even lets you have no hostname and doesn't default to anything! Not that it really matters, but I assume BSD has some way to get the FQDN hostname from hostname.

In terms of curses.h, add an AC_CHECK_HEADERS near https://github.com/OpenLightingProject/ola/blob/0.10/configure.ac#L395 which looks for both curses.h and ncurses/curses.h, then just add some ifdef wrapper around the existing include, and add a new one for the new path, which will come out as HAVE_NCURSES_CURSES_H, so it includes whichever is present and it should just work.

This says you can do a single pipe or for your case statement, which would be preferable if it works:
http://www.delorie.com/gnu/docs/autoconf/autoconf_128.html

This PR can probably target 0.10 again, given it's simply fixing existing functionality on an OS.

[peternewman]

@RenZ0 can you try this commit too (on FreeBSD and DragonFly):
peternewman@11db8a3

The index should stop being -1 and the type should stop being 65535 hopefully on the NetworkTester. I'm assuming FreeBSD is currently broken without the commit, but please double check that too.