Menu
▾
▴
#193 Compiling r7371 and r7372 CSL Reduce under macOS fails.
Compiling native r7371 and r7372 CSL Reduce under arch64 macOS Tahoe 26.5.1 fails!
I didn't try to compile the universal version of CSL Reduce.
As far as r7371 is concerned, the relevant part of the log file is the following:
[aarch64-mac_26_tahoe-darwin25.5.0] CXX csl-arith01.o
In file included from /sw/reduce-algebra/trunk/csl/cslbase/arith01.cpp:41:
In file included from /sw/reduce-algebra/trunk/csl/cslbase/headers.h:83:
In file included from /sw/reduce-algebra/trunk/csl/cslbase/machine.h:317:
/sw/reduce-algebra/trunk/csl/cslbase/float128.h:906:19: error: expected expression
906 | string_to_160(const char* s, sign, x, m);
| ^
1 error generated.
gmake[3]: *** [Makefile:3066: csl-arith01.o] Error 1
gmake[3]: uscita dalla directory «/sw/reduce-algebra/trunk/cslbuild/aarch64-mac_26_tahoe-darwin25.5.0/csl»
gmake[2]: *** [Makefile:1705: all] Error 2
gmake[2]: uscita dalla directory «/sw/reduce-algebra/trunk/cslbuild/aarch64-mac_26_tahoe-darwin25.5.0/csl»
gmake[1]: *** [Makefile:693: /sw/reduce-algebra/trunk/m4/libtool.m4] Error 2
gmake[1]: uscita dalla directory «/sw/reduce-algebra/trunk/cslbuild/aarch64-mac_26_tahoe-darwin25.5.0»
Reduce build tasks finished: Highest return code 2
make: [all] Error 2 (ignored)
The same happens also while compiling under x86_64 macOS Big Sur 11.7.11.
As far as r7372 is concerned, there are also some undefined variables.
The last revision I was able to compile without problems under macOS is r7366.
Best regards
Marco
Hi Arthur.
I'm trying to compile under macOS x86_64 Big Sur 11.7.11 and also under macOS M1 Tahoe 26.5.1.
Compiling and running both r7373 and r7374 native CSL Reduce works fine.
There is however a problem.
Since compiling universal versions of CSL Reduce always fail, the "trunk/macbuild/Makefile" does not work as expected because it will try to compile a universal version of CSL Reduce.
Best regards
Marco
I build a universal version on my Macbook over the weekend and make a
lengthy posting about what was involved - a large fraction of which was
the issue of coaxing macports to get enough installed in universal mode.
Have you read all of what I wrote there? It basically ends up pessimistic
for that task given that Apple are signalling loudly that support for
x86_64 will fade from later this year. So if you need a set of universal
binaries or even just x86_64 ones you need to get your macports world up
to that over the next couple of months. The CSL and Reduce files
themselves will not be wilfully making that harder (a week back I had a
pile of glitches as I cleaned up the code by beiong able to retire us of a
softfloat library, but that should be transient).
Ordinary users are somewhat not expected to use the trunk/macbuild
directory. That is there and used while building snapshots. The
"standard" way to build is to go "./configure --with-csl
[--enable-universal/--disable-universal]" just in the trunk directory.
The Makefile in macbuild does not provide an overridable option re
universal building because when we are about to make a snapshot we either
use a historival set of macports that we managed to get going a while ago
or follow the jolly games as per my posting. Also of course PSL is
defeated my what Apple have done in their recent operating systems, so
the macbuild directory will have pain or be deficient regarding that. The
top level trunk directory separates out ./configure --with-csl and
--with-psl more cleanly. So yes I agree that going "make" in the macbuild
directory is currently not working nicely. Eg on a new macbook it tried to
build a PSL version and that utterly fails because of the incompatible
changes Apple made a few years back. I will look into it at SOME stage but
not with any urgency.
To try to preserve my sanity I have to note that Big Sur was "retired" in
September 2023 and that is now inching along to count as "quite some while
ago". I do not like the way that Apple rendered older but still quite
usable machines obsolete and unsupported (but almost equally hatefully
Microsoft made Windows 11 require new hardware equally trying to force
people to spent money upgrading). So I wish to take the view that I can
not spend much time supporting Big Sur (the pain of keeping old releases
of MacOS in VMs on that platform does not help either), and as of maybe
September or October I am fully expecting that when I upgrade Xcode that I
will cease to be able to compile for x86_64. And it will be something of a
relief to feel able to back off from the effort that univeral binaries
leads to (much of which is the macports situation, but beyoind that
building them takes more elapsed time that compiling for just one
architecture).
Sorry.
Arthur
On Mon, 6 Jul 2026, Marco Ferraris wrote:
Related
Bugs: #193