05.09.2026 12:19, Stuart Henderson пишет:
> On 2026/09/04 20:00, Klemens Nanni wrote:
>> 04.09.2026 22:32, Stuart Henderson пишет:
>>> I think it would be better to set it all the time, like MODSIMPLE_configure
>>> does in bsd.port.mk.
>>
>> Like so?
>
> I meant in python.port.mk, I'd be pretty happy to do that without wider
> testing in advance. Changing it in bsd.port.mk I'm a lot more twitchy
> about..
Sure, let's do the bare minium then wrt. compiler/tool detection:
Index: python.port.mk
===================================================================
RCS file: /cvs/ports/lang/python/3/python.port.mk,v
diff -u -p -r1.28 python.port.mk
--- python.port.mk 19 Jun 2026 13:49:04 -0000 1.28
+++ python.port.mk 5 Sep 2026 15:46:12 -0000
@@ -223,7 +223,7 @@ MODPY_DISTUTILS_BUILD ?= build --build-b
MODPY_DISTUTILS_INSTALL ?= install --prefix=${TRUEPREFIX} \
--root=${DESTDIR}
-MAKE_ENV += CC=${CC}
+MAKE_ENV += CC=${CC} CXX=${CXX}
CONFIGURE_ENV += PYTHON="${MODPY_BIN}"
.if ${CONFIGURE_STYLE:Mgnu}
CONFIGURE_ENV += ac_cv_prog_PYTHON="${MODPY_BIN}" \
OK?
For CXXFLAGS it'd take a closer look when time allows.
>
>
>> I also just noticed -O3 in matplotlib build logs and passing CXXFLAGS
>> make meson append them, so default -O2 is used again and DEBUG gets honoured.
>>
>> bsd.port.mk already adds CFLAGS to MAKE_ENV outside MODSIMPLE_configure,
>> but not CXXFLAGS.
>>
>> How about adding all just there, dropping C* things in python.port.mk?
>>
>> 'make -V MAKE_ENV' in matplotlib shows them all without duplicates from
>> other places.
>>
>>
>> Index: lang/python/3/python.port.mk
>> ===================================================================
>> RCS file: /cvs/ports/lang/python/3/python.port.mk,v
>> diff -u -p -r1.28 python.port.mk
>> --- lang/python/3/python.port.mk 19 Jun 2026 13:49:04 -0000 1.28
>> +++ lang/python/3/python.port.mk 4 Sep 2026 19:54:50 -0000
>> @@ -223,7 +223,6 @@ MODPY_DISTUTILS_BUILD ?= build --build-b
>> MODPY_DISTUTILS_INSTALL ?= install --prefix=${TRUEPREFIX} \
>> --root=${DESTDIR}
>>
>> -MAKE_ENV += CC=${CC}
>> CONFIGURE_ENV += PYTHON="${MODPY_BIN}"
>> .if ${CONFIGURE_STYLE:Mgnu}
>> CONFIGURE_ENV += ac_cv_prog_PYTHON="${MODPY_BIN}" \
>> Index: infrastructure/mk/bsd.port.mk
>> ===================================================================
>> RCS file: /cvs/ports/infrastructure/mk/bsd.port.mk,v
>> diff -u -p -r1.1651 bsd.port.mk
>> --- infrastructure/mk/bsd.port.mk 18 May 2026 23:38:23 -0000 1.1651
>> +++ infrastructure/mk/bsd.port.mk 4 Sep 2026 19:58:38 -0000
>> @@ -754,7 +759,8 @@ PORTHOME ?= /${PKGNAME}_writes_to_HOME
>>
>> MAKE_ENV += PATH='${PORTPATH}' PREFIX='${PREFIX}' \
>> LOCALBASE='${LOCALBASE}' X11BASE='${X11BASE}' \
>> - CFLAGS='${CFLAGS:C/ *$//}' \
>> + CC=${CC} CFLAGS=${CFLAGS:Q} \
>> + CXX=${CXX} CXXFLAGS=${CXXFLAGS:Q} \
>
> have you looked into why it currently strips trailing spaces from
> CFLAGS? could it cause some subtle problem if it stops doing that?
No, the original commit from 2004 just repeats the diff in words
without explaining anything.
I'm inclined to use proper mechanisms like :Q and deal with fallout
when, that is if it should arise.
>
>> TRUEPREFIX='${PREFIX}' ${DESTDIRNAME}='' \
>> HOME='${PORTHOME}'
>>
>>
>
