Introduce stdlib.cat

[hernanponcedeleon]

In herd, most of what we defined in basic.cat is defined in stdlib.cat which is included by default in all models (see section 16.7 in https://diy.inria.fr/doc/herd.html).

This PR mimics this behavior.

[ThomasHaas]

We did not always place the stdlib include at the beginning, for example in vmm.cat. The default include might cause issue in that regard. Well, I guess you can re-include the stdlib.cat after modifying definitions to overwrite the already created one's.

[hernanponcedeleon]

I thought about that, and I still decided to remove it to see if this already causes problems. I assume the vmm.cat from the libvsync repo should not depend on another cat files so this probably will need to be fixed in the official model. I will discuss that with jonas.

[ThomasHaas]

The failed checks even include the RelationAnalysisTest. That means you just exposed some more bugs in RA or AA I think.

[hernanponcedeleon]

The failed checks even include the RelationAnalysisTest. That means you just exposed some more bugs in RA or AA I think.

Maybe it is the non-det from #746 and this was just an unlucky run.

[ThomasHaas]

But the non-det was supposed to only affect the new alias analysis which is disabled inRelationAnalysisTest. So if it is non-det behaviour, then you exposed it on the older versions of AA.

[hernanponcedeleon]

I thought about that, and I still decided to remove it to see if this already causes problems. I assume the vmm.cat from the libvsync repo should not depend on another cat files so this probably will need to be fixed in the official model. I will discuss that with jonas.

The latest version of vmm already solves most of the issues, it only need to properly define po-loc

EDIT: we don't need to redefine po-loc since it does not depends on the new ext and int

[ThomasHaas]

LGTM