From: Landry Breuil <landry@openbsd.org>
Subject: Re: [maintainer update] hdf5 1.14.6
To: Theo Buehler <tb@theobuehler.org>
Cc: Martin Reindl <martin@catai.org>, ports@openbsd.org
Date: Thu, 4 Sep 2025 14:55:20 +0200

Le Thu, Sep 04, 2025 at 02:43:25PM +0200, Landry Breuil a écrit :
> Le Wed, Sep 03, 2025 at 10:04:06PM +0200, Theo Buehler a écrit :
> > On Wed, Sep 03, 2025 at 09:39:00PM +0200, Martin Reindl wrote:
> > > On Wed, Sep 03, 2025 at 07:09:33PM +0200, Theo Buehler wrote:
> > > > On Wed, Aug 27, 2025 at 08:07:07PM +0200, Martin Reindl wrote:
> > > > > This diff updates hdf5 to 1.14.6 an enables the use of libaec.
> > > > 
> > > > [...]
> > > > 
> > > > > There might be more ports which start to pick up libaec leading
> > > > > to possible conflicts with szlib, so I am looking for OKs to
> > > > > this update.
> > > > 
> > > > I don't understand what you're saying here. With what and how do you
> > > > expect libaec or szlib to conflict?
> > > > 
> > > > I started a bulk before looking at your diff and qgis,qt6 broke with
> > > > "ld.so: crssync: can't load library 'libaec.so.0.0'" because libhdf5
> > > > was installed, but not libaec. The problem is this:
> > > > 
> > > > > +BUILD_DEPENDS +=	archivers/libaec
> > > > 
> > > > This should be a LIB_DEPENDS and you need WANTLIB += aec sz.
> > > > (make port-lib-depends-check hints at that)
> > > > 
> > > > This may result in several dependent ports requiring a wantlib sync.
> > > > I have not checked that.
> > > > 
> > > > Purely from a symbol perspective only hdf5 and hdf5_fortran need a major
> > > > bump, but given the library interdependencies in this port, the major
> > > > bump for all shared libs is the right thing to do.
> > > 
> > > I sent out an old version of the diff in the original message. Here is
> > > the correct one.
> > 
> > That looks better, thanks. It matches what is in my bulk now.
> > 
> > Unless I run into more issues in my bulk, I think this should basically
> > be good to go.
> > 
> > We still need to know what consumers need a wantlib adjustment and
> > landry wants to check the impact on the geo/ side of things, hopefully
> > this week-end.
> 
> i've had a quick look and built hdf5 with the last diff (with
> LIB_DEPENDS pointing at archivers/libaec) and gdal/mdal/postgis built
> fine, they need WANTLIB tweaks but i'll handle them, no worries.
> 
> ok with me too.

minor nit, but your --with-szlib=${PREFIX}/include,${PREFIX}/lib
CONFIGURE_ARGS looked weird to me, was there a specific reason for that ?

it feels cleaner/more standard using --with-szlib=${PREFIX}, and
configure detects it fine:

checking for SZ_BufftoBuffCompress in -lsz... yes
checking for szlib.h... yes
checking for szlib encoder... yes

same for FC=egfortran, even if it isnt changed in your diff, should that
be a CONFIGURE_ENV variable ? it feels weird in CONFIGURE_ARGS.

Landry