Hi, I didn't expect it to be that much more work, Now seeing that, the initial version seems to be the better tradeoff.
Sebastian On Wed, Mar 4, 2026 at 1:03 AM Stuart Henderson <[email protected]> wrote: > On 2026/03/03 22:06, Theo Buehler wrote: > > > here goes. on the one hand, renaming the library makes it significantly > > > more hassle to patch ports to use it, on the other it no longer needs > to > > > go in a subdir so no more rpath, and "make port-lib-depends-check" > works. > > > > I can live with this, but I think it's not a great trade-off. I would > have > > preferred the original eopenssl-like approach. > > > > for an example of what "significantly more hassle" looks like, here's > how patches for the two approaches look for nmap (which doesn't use > pkg-config for this). > > $ acvs di patches|filterdiff -i '*_ac'|diffstat > patch-configure_ac | 12 ------------ > 1 file changed, 12 deletions(-) > > $ acvs di patches|filterdiff -i '*_ac'|diffstat > patch-configure_ac | 46 > ++++++++++++++++++++++++++++++++++++++-------- > patch-ncat_configure_ac | 39 > +++++++++++++++++++++++++++++++++++++++ > patch-nping_configure_ac | 39 > +++++++++++++++++++++++++++++++++++++++ > patch-nsock_src_configure_ac | 26 ++++++++++++++++++++++++++ > 4 files changed, 142 insertions(+), 8 deletions(-) > > it might be possible to trim that down with more work, but it took > a while to get there. for the "same name different dir" approach it's > a simpler case of adding "-L${LOCALBASE}/lib/libpcap > -Wl,-rpath,${LOCALBASE}/lib/libpcap" to LDFLAGS, and the only thing > which tripped me up there was when I typoed as ${LOCALBASE}/libpcap > and it took me a few rounds to notice what was up. > >
