Quentin Sabah

I'll try to reproduce the issue manually (by invalidating a repo key) and test the alternatives, then write a PR.

Two tests fails with segfault on Windows 2025 CI, `magic_nqueens` and `inline_nqueens`. I don't have access to a Windows + MSCV dev env, so I cannot investigate. Will open issues...

I confirm. The identifier of the program is prepared here: https://github.com/souffle-lang/souffle/blob/5682a9f12e2668ecdd26348fe63cc508bc0fcf47/src/MainDriver.cpp#L1110 `identifier()` replaces the `-` by `_` because it must be a valid C++ identifier for the C++ code generator....

Hi, it's a limitation, the provenance explainer does not display ADT values yet. https://github.com/souffle-lang/souffle/blob/5682a9f12e2668ecdd26348fe63cc508bc0fcf47/src/include/souffle/provenance/ExplainProvenance.h#L231-L254 I believe we could easily reuse or adapt display code from `WriteStream::outputADT` and `WriteStream::outputRecord` to support...

Hi, thanks for reporting this. AFAIK the datalog output has been designed for human review and debugging, hence this kind of discrepancies may remain here and there despite some efforts...

The failing tests in Windows is a known issue since Github runners switched to Windows 2025: https://github.com/souffle-lang/souffle/issues/2548

Hi, if valid dates must be of length 8, my suggestion is the following: create a relation `valid_date(d) :- date(d), strlen(d) = 8.` and then use `valid_date` instead of `date`....

In general the best solution would be to implement multi-result user-defined functors: functors that would generate zero, one or many results, very much like relation atoms. This is partly discussed...