Re: py313-pybind11 build fails on netbsd-11/amd64?

["John D. Baker" <jdbaker%consolidated.net@localhost>]

On Sat, 14 Mar 2026, Greg Troxel wrote:

> "John D. Baker" <jdbaker%consolidated.net@localhost> writes:
> 
> > Having updated to pkgsrc-HEAD prior to the freeze leading to pkgsrc-2026Q1,
> > I've updated my usual complement of packages for netbsd-11 (11.0_RC2) for
> > NetBSD/amd64 except for "graphics/py313-Pillow" because its dependency,
> > "devel/py313-pybind11" fails to build as follows:
> 
> What architecture?  Are you using i386?

This is amd64 (x86_64):

$ uname -a
NetBSD plex760.technoskunk.fur 11.0_RC2 NetBSD 11.0_RC2 (PLEX760) #19: Wed Mar  4 22:06:19 CST 2026  sysop%plex760.technoskunk.fur@localhost:/r0/build/netbsd-11/obj/amd64/sys/arch/amd64/compile/PLEX760 amd64

> I today built pybind11 on NetBSD 10 amd64 and I just built it on NetBSD
> 11 (11.0_RC2) amd64.
> 
> I am confused that pybind11 is depending on scikit.

$ pwd 
/x/pkgsrc/devel/py-pybind11
$ make show-all-depends
depends:
  pkg   TOOL_DEPENDS=           \
                                cmake>=3.15:../../devel/cmake \
                                cwrappers>=20150314:../../pkgtools/cwrappers \
                                mktools-[0-9]*:../../pkgtools/mktools \
                                py313-build>=0:../../devel/py-build \
                                py313-installer>=0.7.0nb1:../../misc/py-installer \
                                py313-scikit-build-core>=0.11.2:../../devel/py-scikit-build-core \
                                # end of TOOL_DEPENDS (sorted)
  pkg   BUILD_DEPENDS=          # empty
  pkg   TEST_DEPENDS=           \
                                py313-test-[0-9]*:../../devel/py-test \
                                # end of TEST_DEPENDS (sorted)
  pkg   DEPENDS=                \
                                python313>=3.13.0:../../lang/python313 \
                                python313>=3.13:../../lang/python313 \
                                # end of DEPENDS (sorted)
  def   _DEPENDS_FILE=          /tmp/pkgsrc/devel/py-pybind11/work/.depends
  def   _RDEPENDS_FILE=         /tmp/pkgsrc/devel/py-pybind11/work/.rdepends


 
> I just built py313-Pillow (N 11 amd64) and that went fine, and had no
> deps on numpy/scikit/scipy, just graphics format libs.  That was on
> 2025Q4.  It builds fine on pkgsrc-HEAD on N 10 amd64.

$ pwd
/x/pkgsrc/graphics/py-Pillow
$ make show-all-depends
depends:
  pkg   TOOL_DEPENDS=           \
                                cwrappers>=20150314:../../pkgtools/cwrappers \
                                glib2-tools-[0-9]*:../../devel/glib2-tools \
                                mktools-[0-9]*:../../pkgtools/mktools \
                                pkgconf-[0-9]*:../../devel/pkgconf \
                                py313-build>=0:../../devel/py-build \
                                py313-installer>=0.7.0nb1:../../misc/py-installer \
                                py313-setuptools>=78:../../devel/py-setuptools \
                                # end of TOOL_DEPENDS (sorted)
  pkg   BUILD_DEPENDS=          \
                                py313-pybind11>=2.5.0:../../devel/py-pybind11 \
                                # end of BUILD_DEPENDS (sorted)
  pkg   TEST_DEPENDS=           \
                                ghostscript-[0-9]*:../../print/ghostscript \
                                netpbm-[0-9]*:../../graphics/netpbm \
                                py313-numpy-[0-9]*:../../math/py-numpy \
                                py313-olefile-[0-9]*:../../devel/py-olefile \
                                py313-test-[0-9]*:../../devel/py-test \
                                py313-test-cov-[0-9]*:../../devel/py-test-cov \
                                py313-test-timeout-[0-9]*:../../devel/py-test-timeout \
                                # end of TEST_DEPENDS (sorted)
  pkg   DEPENDS=                \
                                freetype2>=2.1.3:../../graphics/freetype2 \
                                freetype2>=2.13.0:../../graphics/freetype2 \
                                freetype2>=2.13.2nb1:../../graphics/freetype2 \
                                lcms2>=2.2:../../graphics/lcms2 \
                                libavif>=1.0.3:../../graphics/libavif \
                                libimagequant>=4.2.0:../../graphics/libimagequant \
                                libjpeg-turbo>=1.1.0:../../graphics/libjpeg-turbo \
                                libwebp>=0.2.0:../../graphics/libwebp \
                                libwebp>=0.6.0:../../graphics/libwebp \
                                openjpeg>=2.1.0:../../graphics/openjpeg \
                                openjpeg>=2.1.0:../../graphics/openjpeg \
                                python313>=3.13.0:../../lang/python313 \
                                python313>=3.13:../../lang/python313 \
                                raqm>=0.10.1:../../graphics/raqm \
                                raqm>=0.10.3nb1:../../graphics/raqm \
                                tiff>=3.6.1:../../graphics/tiff \
                                tiff>=4.7.0nb3:../../graphics/tiff \
                                # end of DEPENDS (sorted)
  def   _DEPENDS_FILE=          /tmp/pkgsrc/graphics/py-Pillow/work/.depends
  def   _RDEPENDS_FILE=         /tmp/pkgsrc/graphics/py-Pillow/work/.rdepends


> Are you sure your tree matches pkgsrc HEAD?

It's been a day or so since I last updated, but I can be sure to
use "-A" again the next time to be sure there are no tags from
pkgsrc-2025Q4 hanging around.

If that doesn't work, move "distfiles" and "wip" aside, nuke it and check
it out again from scratch.

> As for the mismatch, on i386 we build for i486.  But there are cpuflags
> sometimes, and there's atomic64.mk, and that might end up with a higher
> cpu type that might enable some vector stuff, and then some other lib
> might not use atomic64 and not use that same vector stuff.  That's some
> blend of speculation, rambling, and a clue; don't take it too seriously.

I'm not (yet) building anything for i386(i486), but when I do, this may
become relevant.

-- 
|/"\ John D. Baker, KN5UKS               NetBSD     Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSD            FreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645