On 2019/03/12 20:44, Klemens Nanni wrote:
> WANTLIB and USE_GMAKE have a space after "=".
> Please break LIB_DEPENDS into one port per line.

+1

It gained a BUILD_DEPENDS on gtk2mm, but it's only actually used in
one of the examples (disabled via CONFIGURE_ARGS) and they don't actually
build (for another reason) if you enable them.

I think I would drop the gtk2mm dependency; as-is, if it changes to being
a library dependency rather than just a build dependency in the future,
it wouldn't be noticed in bulk builds.


> You can set SEPARATE_BUILD.
> 
> As per the project's README, `--enable-debug' should be passed for extra
> "debugging code", so you could add that to CONFIGURE_ARGS if DEBUG is
> set in case anyone cares to hack on using this port.
> 
> On Wed, Mar 06, 2019 at 11:55:17AM +0100, Stefan Sperling wrote:
> > 1) Add more patches, and add comments to all patches.
> Thanks, appreciated.
> 
> >    In particular, fix poll(2) event mask which resulted in unhandled
> >    POLLHUP events in the main event loop which made textsuggest-server
> >    spin on the CPU. As far as I understand, the "unlock" file descriptors
> >    are only interested in POLLIN events. They can be triggered to wake the
> >    event loop while a thread sleeps in poll(2). This feature is used by the
> >    library's teardown code. POLLOUT makes no sense here.
> >    textsuggest uses much less CPU becomes a lot more responsive with this 
> > fix.
> Do you intend to push those upstream?
> 
> > 2) Enable glib integration. This adds another shared library which is not
> >    used by textsuggest but will perhaps be useful if other consumers of
> >    libdbus-c++ show up.
> > 
> > 3) Run regression test (there's just one, the others are just empty stubs).
> 
>       $ make test
>       ===>  Regression tests for libdbus-c++-0.9.0
>       
> /usr/ports/pobj/libdbus-c++-0.9.0/libdbus-c++-0.9.0/test/functional/Test1/TestApp
>       initialize DBus...
>       terminating with uncaught exception of type DBus::Error: 
> /usr/local/bin/dbus-launch terminated abnormally with the following error: No 
> protocol specified
>       Autolaunch error: X11 initialization failed.
> 
>       *** Signal SIGABRT in . (Makefile:31 'do-test')
> 
> > Still OK?
> Assuming tests are an error on my site, OK kn.
> 

Tests fail with PORTS_PRIVSEP but run successfully otherwise (this is
common for most things using X in tests, we don't have a good way to fix
that so that can be ignored in this port).


Reply via email to