execution on reals make CI fail
I realized this morning that after merging !647 (merged) into master, the CI of master failed: https://gitlab.inria.fr/why3/why3/-/pipelines/476002/failures
This occurs in a test of execution on real numbers, the interpreter runs out of memory.
I can't see why the changes of !647 (merged) could have anything to do with execution on real numbers.
I can't reproduce the bug on my laptop, my opam won't want to install mlmpfr apparently for compatibility mismatch on libmpfr.
Looking at the test that fails, I wonder if we could workaround the error by changing the non binary-representable literal 0.002
into say 0x1p-10
. But I can't test that.
Any suggestion welcome ! (@melquion ?)