NEWS: move from CVC4 to cvc5
Makarius
makarius at sketis.net
Fri Jan 31 15:46:57 CET 2025
*** HOL ***
* Sledgehammer:
- Update of bundled provers:
. cvc5 1.2.0 -- with support for arm64-linux
This refers to Isabelle/bbda3b4f3c99.
That is not just an update from version 4 to 5, but the change of spelling the
name indicates that it is a quite different system.
One guy at Lean Together 2025, when talking about Sledgehammer and
Magnushammer even put an exclamation mark at "Use cvc5!" (see minute 4:25 of
https://www.youtube.com/watch?v=T9fhMJmJRAw&list=PLlF-CfQhukNlzXdQvu1SVt9vcD4--fLlg&index=20).
The support for arm64-linux was not available before, which is extra
motivation to move on now. Thus we are almost complete in supporting that
platform.
Testing cvc5 beforehand, we did encounter a few oddities concerning signals
already, but lets hope that this workaround is now sufficient:
changeset: 81746:377887fbc968
user: wenzelm
date: Wed Jan 08 14:30:17 2025 +0100
files: Admin/components/components.sha1 src/Pure/Admin/component_cvc5.scala
description:
rebuild cvc5 component (still inactive);
more robust workaround on all platforms, avoid crash of "sledgehammer [cvc5]"
seen on x86_64-linux after line 75 of
"$AFP/thys/Boolos_Curious_Inference_Automated/Boolos_Curious_Inference_Automated.thy";
It wraps everything into an outermost bash process, and refrains from using
"exec" to make a "POSIX tail call". Thus the resulting process hierarchy looks
more like a regular Terminal / shell process. (There are many fine points on
process hierarchies and signals that can be done wrong. While I don't
understand everything myself, I've seen enough bad approaches to be avoided.)
Makarius
More information about the isabelle-dev
mailing list