On Sun, Nov 06, 2016 at 06:33:52PM -0700, Andrew Fresh wrote:
> I was trying to build p5-DBD-SQLite as part of the process of updating
> p5-DBIx-Class and I couldn't get it to work.
> 
> If you currently
> cd /usr/ports/databases/p5-DBD-SQLite && make test
> 
> It fails with this, ever so helpful, message:
> 
> perl:/mnt/memory/pobj/p5-DBD-SQLite-1.50/DBD-SQLite-1.50/blib/arch/auto/DBD/SQLite/SQLite.so:
>  undefined symbol 'sqlite3_column_database_name'
> perl:/mnt/memory/pobj/p5-DBD-SQLite-1.50/DBD-SQLite-1.50/blib/arch/auto/DBD/SQLite/SQLite.so:
>  undefined symbol 'sqlite3_column_table_name'
> Can't load 
> '/mnt/memory/pobj/p5-DBD-SQLite-1.50/DBD-SQLite-1.50/blib/arch/auto/DBD/SQLite/SQLite.so'
>  for module DBD::SQLite: Cannot load specified object at 
> /usr/libdata/perl5/amd64-openbsd/5.20.3/DynaLoader.pm line 193.
> 
> 
> This diff at least make *most* of the p5-DBD-SQLite tests pass, but
> t/43_fts3.t test won't seem to pass, no matter what I enable.
> 
> Obviously the comments would go away, but thought they might be useful
> if someone else has some idea why it isn't working.  I may have time to
> dig into it more in the next week, but going out to dinner now.
> +
> +# I also tried adding all of this, from when it was in src
> +# but that didn't help with the test either.
> +# 
> http://cvsweb.openbsd.org/cgi-bin/cvsweb/src/lib/libsqlite3/Attic/Makefile?rev=1.16&content-type=text/x-cvsweb-markup
> +# CFLAGS +=  -DSQLITE_ENABLE_COLUMN_METADATA \
> +#            -DSQLITE_ENABLE_RTREE \
> +#            -DSQLITE_ENABLE_UNLOCK_NOTIFY \
> +#            -DSQLITE_ENABLE_FTS3 \
> +#            -DSQLITE_ENABLE_FTS3_PARENTHESIS \
> +#            -DSQLITE_ENABLE_DBSTAT_VTAB \
> +#            -DSQLITE_OMIT_BUILTIN_TEST
> +#CFLAGS +=   -DSQLITE_THREADSAFE=1 \


Ooops. All those ones should be re-enabled nevertheless, i know mozilla
depends on at least unlock_notify and fts3 if we want to run properly
against systemwide sqlite, see

http://hg.mozilla.org/releases/mozilla-release/file/tip/old-configure.in#l4580

i cheat for SECURE_DELETE since there's a patch doing this via a pragma
at runtime, but the others are configure-time checked.

Reply via email to