On Fri, Oct 30, 2015 at 10:02:07AM +0000, Stuart Henderson wrote:
> On 2015/10/30 09:40, Landry Breuil wrote:
> > On Thu, Oct 29, 2015 at 08:24:08PM -0700, patrick keshishian wrote:
> > > On 10/29/15, patrick keshishian <[email protected]> wrote:
> > > > On 10/29/15, Landry Breuil <[email protected]> wrote:
> > > [snip]
> > > >> Now that postgresql.port.mk allows to run those tests with the correct
> > > >> charset, can we get an updated tarball with a patch for the default
> > > >> --cache value (and the corresponding discussion brought upstream) ? I'd
> > > >> like to import this one this weekend....
> > > 
> > > Per request attached is a new tar-ball that takes more tests
> > > from FAIL to PASS. test-suite.log also attached.
> > 
> > Nice... some nits:
> > 
> > - patches will need comments, the ones fixing %format issues like
> >   patch-node-ram-cache_cpp need to go upstream in a PR, as
> > patch-sprompt_cpp and patch-tests_common-pg_cpp
> 
> The %zd -> %llu may break some other OS, it may need a cast if
> going upstream.

Just fixed upstream by
https://github.com/openstreetmap/osm2pgsql/pull/483, which was proposed
just before mine in https://github.com/openstreetmap/osm2pgsql/pull/487
:)

> > - patch-node-persistent-cache_cpp puzzles me :) a -Werror case ?
> 
> looks like it.

Builds fine without it, anyway i'll push it upstream too.

> > I'll try to set cache to 300 directly as a default value inside
> > osm2pgsql instead of setting it for each and every test, so that by
> > default it works on OpenBSD with default system limits. I'll also
> > look into which limits are hit so that it can be documented.
> > 
> > As for the new cmake-based build, since there's been no release with it
> > yet, that can be worked on later...

Fwiw the new cmake build system works for me:

mkdir /usr/obj/osm2pgsql && cd /usr/obj/osm2pgsql
env CXX=eg++ CXXFLAGS=-pthread cmake path/to/src/osm2pgsql/
make

builds fine.

> | WANTLIB +=          boost_filesystem boost_system boost_system-mt
> | WANTLIB +=          boost_thread-mt bz2 c crypto geos iconv lzma
> | WANTLIB +=          m pq proj protobuf-c pthread ssl stdc++ xml2 z
> 
> no need for iconv, it comes from the module
> 
> | MODULES =           converters/libiconv     \
> |                     databases/postgresql    \
> |                     gcc4                    \
> |                     lang/lua                \
> 
> I think the \ at the end of setting variables is a bit dangerous.
> Probably wants forcing the lua version in case the default system
> version isever changed, e.g. MODLUA_VERSION=5.1
> 
> | CONFIGURE_ENV +=    LUA=lua51
> 
> MODLUA_BIN
> 
> | CONFIGURE_ENV +=    LUA_INCLUDE=-I${LOCALBASE}/include/lua-5.1
> 
> MODLUA_INCL_DIR
> 
> | CONFIGURE_ENV +=    LUA_LIB=-llua5.1
> 
> MODLUA_LIB
> 
> Also set MODLUA_SA=Yes.
> 
> |     @${SUBST_CMD} ${WRKSRC}/docs/osm2pgsql.1
> 
> The patch just has "/usr/local" rather than TRUEPREFIX, probably from
> running update-patches after building. Personally I'd leave the patch like
> it is and change the SUBST_CMD to sed -i s,/usr/local,${TRUEPREFIX}, to
> make updates easier, but otherwise the patch needs adjusting.

Will fix all that, and also add lang/python module for MODPY_ADJ_FILES
which seems needed for tests/regression-test.py. I still have some
issues with tests running with default limits, even when force-setting
cache to 100Mb..

Landry

Reply via email to