On Fri, Sep 18, 2015 at 11:50:19PM -0700, patrick keshishian wrote: > On 9/17/15, Landry Breuil <[email protected]> wrote: > > On Sat, Sep 12, 2015 at 09:05:10AM -0700, patrick keshishian wrote: > >> On 9/12/15, Daniel Jakots <[email protected]> wrote: > >> > On Fri, 11 Sep 2015 16:45:31 -0700, patrick keshishian > >> > <[email protected]> wrote: > >> > > >> >> Does the Makefile look better? I think it is headed toward a more > >> >> correct direction. > >> > > >> > Hi, > >> > > >> > I tried to build it, I had an error, but Patrick quickly sent me a new > >> > Makefile (attached), I could build it without problem then. I imported > >> > a > >> > bunch of osm data in a pgsql database to test, I had the same output > >> > that I usually have on Ubuntu so I guess it's good. > >> > > >> > Thanks for the port! > >> > > >> > Cheers, > >> > Daniel > >> > >> Thanks for your test and report Daniel. > >> > >> Attached the new tar-ball. > > > > After actually looking into it... > > > > - you have a typo in the patch adding without-lua to configure.ac... but > > since this wasnt a build option provided by upstream, why bother with > > adding it and doing a flavor for this? i mean, lua dependency isnt > > usually big, and since it provides quite some useful features > > (scriptable tag transforming ?) why not enabling it by default ? > > > > - REVISION should be removed, no need to start at 0 > > > > - devel/gmake doesnt need to got to BDEP, it's usually done with > > USE_GMAKE=Yes > > > > - make check fails here: > > tests/middle-tests.cpp:7:17: error: tuple: Fichier ou répertoire > > introuvable > > tests/middle-tests.cpp: In function 'int test_node_set(middle_t*)': > > tests/middle-tests.cpp:54: warning: comparison between signed and unsigned > > integer expressions > > tests/middle-tests.cpp: In function 'int > > test_nodes_comprehensive_set(middle_t*)': > > tests/middle-tests.cpp:79: error: 'class expected_nodelist_t' has no member > > named 'emplace_back' > > > > Tried installing libpqxx (as a potential candidate for providing tuple > > header) but that didnt help... installing g++ 4.9 didnt help either. > > Did you run the tests ? > > Your comment one libpqxx threw me for a wild-goose chase. > The intended <tuple> is from c++/4.9.3/tuple. > > Attached is new tar-ball that compile, but unfortunately "make test" > mostly fails[1]. > > Most failures are due test DB not existing followed by core dump. > One hints of bad option. > One on Python module import. > > I figure the last two should be easy to fix. The ones with DB not found > failures might be a bit tougher to figure out. >
Looking a bit more at the new version of the Makefile itself: CONFIGURE_ENV += CC=egcc CXX=eg++ CPP=ecpp This shouldnt be needed, using gcc4 module properly sets symlinks in WRKDIR/bin so that they are used by default. AUTO_ENV isnt a known/used idiom, and it seems strange to define this just to use it once... If the port itself builds fine with base toolchain, and only the testsuite needs recent g++ headers, that's a bit sad to have the whole thing being built with ports gcc ... maybe set NO_TEST with a comment ? Fix the offending code using tuple header ? dunno. Landry
