On 2026/05/21 07:40, Theo Buehler wrote:
> I think this would have helped avoid a recent mistake I made in
> games/freeorion.
> 
> boost_python${MODPY_MAJORMINOR} aka boost_python${MODPY_VERSION:C/\.//g}
> could perhaps also be handled, but before I spend time on figuring out
> how let's see if there's a reason not to do the simple thing:
> 
> Index: python.port.mk
> ===================================================================
> RCS file: /cvs/ports/lang/python/3/python.port.mk,v
> diff -u -p -r1.25 python.port.mk
> --- python.port.mk	16 Mar 2026 22:28:04 -0000	1.25
> +++ python.port.mk	15 May 2026 10:12:23 -0000
> @@ -255,6 +255,9 @@ SUBST_VARS :=	MODPY_PYCACHE MODPY_ABI3SO
>  		MODPY_BIN MODPY_DISTV MODPY_VERSION \
>  		MODPY_PY_PREFIX MODPY_PYOEXTENSION ${SUBST_VARS}
>  
> +CHECK_LIB_DEPENDS_ARGS += \
> +	-S MODPY_WANTLIB="${MODPY_WANTLIB}"
> +
>  UPDATE_PLIST_ARGS += \
>  	-S MODPY_PYOEXTENSION
>  
> 

That looks right and works with what I've tested (and fixes the most
annoying cases where "remove old WANTLIB, check, paste in
check-lib-depends output" puts back the hardcoded version).  ok.

I'm not so bothered about MODPY_MAJORMINOR, it doesn't affect too many
ports, and quite a few of ports affected by it just use it in plists
(e.g. libsamba-python.cpython-314-private-samba.so) where we'd either
need quite a few variants, or have a lot of false positives.