Re: Boost update to 1.89.0

[Thomas Klausner <wiz%netbsd.org@localhost>]

On Thu, Sep 25, 2025 at 09:59:25PM +0100, Adam Ciarciński wrote:
> It took me substantially long time, but finally I've got Boost ready to be updated to 1.89.0. It has one important change that might break several packages: boost_system shared library has been removed and the boost_system component is only available as header files. In practice, a package might try to detect the presence of boost_system library or try to link against it. The simple work-around is to remove any reference to boost_system from linker flags and configure scripts.

Thanks, good to know.

> I've tested meta-pkgs/bulk-test-boost and, besides for the boost update, I have the patches and fixes for these packages:
> 
> * botan2: patch + update to 2.19.5
> * gnucash: patch
> * convertes/librevenge: remove patches/patch-librevenge.pc.in
> * graphics/pcl: patch
> * archivers/innoextract: patch
> * math/libixion: patch
> * graphics/gource: patch
> * ham/gnuradio-core: patch
> * net/icinga2: patch + update to 2.15.0
> * misc/libreoffice: remove LDFLAGS from main Makefile
> * audio/ncmpcpp: update to 0.10.1
> * geography/osm2pgsql: patch
> * cad/openscad: patch
> 
> I wasn't able to build the following for reasons not related to Boost:
> 
> - blender: failes for undefined 'ustringHash' - I have no idea what defined that in the past (maybe openimageio?), but the definition is gone (I know it should be an alias to uint64_t); clearly, Blender needs an update.

That matches bulktracker - seems it broke sometime between July and September.

> - devel/libftdi1: doesn't build because of swig - tried Python 3.12 and Python 3.13 without success.

This package builds fine for me on 11.99.2/x86_64.
What symptoms do you see?

> - misc/libreoffice: fails to build on my NetBSD-aarch64 with "! Error: CFI instruction used without previous .cfi_startproc".

bulktracker is unclear on that one. We have a successful build for
NetBSD 10.0/aarch64 but a failure for NetBSD 10.0/aarch64eb, but both
for the 25.2.4.3 version from 2025Q2; no results for the current
version yet; and failures for Darwin/aarch64*.

> If there are no objections, I would like to proceed with the changes anytime soon.

No objections from me, though I'd like some details for libftdi1 first.

Thanks,
 Thomas