Build problem on Mac OS

[egri-nagy]

Semigroups packages fails to build on Mac OS 15.5.
Compiler:

Apple clang version 17.0.0 (clang-1700.0.13.5)
Target: arm64-apple-darwin24.5.0
Thread model: posix

Error message:

 make
./config.status gen/pkgconfig.h
config.status: creating gen/pkgconfig.h
config.status: gen/pkgconfig.h is unchanged
echo > gen/pkgconfig.h.stamp
   GAC     src/bipart.cpp => gen/src/bipart.o
   GAC     src/cong.cpp => gen/src/cong.o
   GAC     src/conglatt.cpp => gen/src/conglatt.o
   GAC     src/froidure-pin-base.cpp => gen/src/froidure-pin-base.o
   GAC     src/froidure-pin-bipart.cpp => gen/src/froidure-pin-bipart.o
   GAC     src/froidure-pin-bmat.cpp => gen/src/froidure-pin-bmat.o
   GAC     src/froidure-pin-fallback.cpp => gen/src/froidure-pin-fallback.o
   GAC     src/froidure-pin-matrix.cpp => gen/src/froidure-pin-matrix.o
   GAC     src/froidure-pin-max-plus-mat.cpp => gen/src/froidure-pin-max-plus-mat.o
   GAC     src/froidure-pin-min-plus-mat.cpp => gen/src/froidure-pin-min-plus-mat.o
   GAC     src/froidure-pin-pbr.cpp => gen/src/froidure-pin-pbr.o
   GAC     src/froidure-pin-pperm.cpp => gen/src/froidure-pin-pperm.o
   GAC     src/froidure-pin-transf.cpp => gen/src/froidure-pin-transf.o
   GAC     src/pkg.cpp => gen/src/pkg.o
In file included from src/pkg.cpp:59:
In file included from ./bin/include/libsemigroups/sims1.hpp:1269:
./bin/include/libsemigroups/sims1.tpp:622:39: error: no member named 'long_rules' in 'libsemigroups::Sims1::iterator_base'
  622 |     _longs           = std::move(that.long_rules());
      |                                  ~~~~ ^
1 error generated.
make: *** [gen/src/pkg.o] Error 1

libsemigroups 3.0.3 by itself compiles without problem.

Being discussed on the GAP Forum https://lists.uni-kl.de/gap/arc/forum/2025-06/msg00008.html

[james-d-mitchell]

Thanks for the report @egri-nagy. This is the same issue as reported at:

libsemigroups/libsemigroups#557

There are 3 solutions (from quickest to slowest):

1. The quickest fix (if you just want to get Semigroups to compile) is to just delete that line of code.
2. The next best fix will be that I release a new version of libsemigroups version 2, which fixes the issue (the stable-2.7 branch of libsemigroups already contains the fixed version of the code). Then release a new version of Semigroups using the new release of libsemigroups.
3. The best fix would be that I update the Semigroups package to work with version 3 of libsemigroups.

I will do 2 as soon as I reasonably can (probably not this week), and am working on 3.

[dimpase]

the current master of Semigroups still has things like hardcoded std=gnu++14, while you are using gnu++17 or even gnu++23 features. (actually, same applies to libsemigroups, so you need them to be in sync here, I guess)

This doesn't fly with new Apple's clang 17.

[james-d-mitchell]

while you are using gnu++17 or even gnu++23 features.

We are not using c++17 or c++23 features in Semigroups at the moment.

(actually, same applies to libsemigroups, so you need them to be in sync here, I guess)

libsemigroups version 3 is c++17, and does not use any features from later c++ versions. Libsemigroups version 2 (which is required by the Semigroups package for GAP) requires only C++14, so I don't think your comment is related to this issue @dimpase.

[dimpase]

@james-d-mitchell I tried to see if the unreleased libsemigroups (3.0.3-...) would work with Semigroups and clang 17. It doesn't - Semigroups is still c++14, it seems.

By the way, are you considering to move from fmt to std::format ? At least in c++23 the latter has everything you use in fmt, IMHO

[egri-nagy]

1. The quickest fix (if you just want to get Semigroups to compile) is to just delete that line of code.

For anyone in a hurry, I can confirm that the quick fix works fine.

[james-d-mitchell]

Resolved in v5.5.1 but I'm going to leave this issue open until that's picked up by a GAP release itself (just in case anyone else has this problem).

[dimpase]

what are the versions of libsemigroups which are compatible with 5.5.1?

[james-d-mitchell]

Good question, only v2.7.4 at the moment

[dimpase]

not v3 at all, right?

[james-d-mitchell]

not v3 at all, right?

Right. We're working on it but it might take a little while.