Skip to content

Memory "softcoding": regression on equivalences #819

Closed

Description

@loutr

Regression introduced by commit 23cf445. MWE:

type t.

module M = {
  var tf, tg : t 

  proc f() = {
    var x;
    x <- tf;
    return x;
  }

  proc g() = {
    var x;
    x <- tg;
    return x;
  }
}.

equiv fg : M.f ~ M.g : M.tf{1} = M.tg{2} ==> ={res}.
proc. (* anomaly: Failure("expecting memory") *)

Replacing the function bodies by return tf;/return tg; makes the proc tactic go through.

I'll try to investigate if I have some time. But I believe this is not the only regression introduced by this commit.

Activity

  1. fdupress commented on Oct 14, 2025

    @fdupress
    Member

    Please add to #818 and close this as duplicate?

    And consider adding the repos containing the regressions to external CI.

  2. loutr commented on Oct 14, 2025

    @loutr
    ContributorAuthor

    Is it really a duplicate though? At least the output anomaly is not the same. I did not even know there was a repo with regression snippets. I will try to have a look and change the CI but this bug caught me while I was doing something else that I would like to finish first.

  3. fdupress commented on Oct 14, 2025

    @fdupress
    Member

    I did not even know there was a repo with regression snippets.

    There is not, but I assumed you had a proof going that triggered this. You can add that repo to the external CI. (Which caught quite a few regressions before we merged already.)

  4. fdupress commented on Oct 14, 2025

    @fdupress
    Member

    Is it really a duplicate though? At least the output anomaly is not the same.

    You are very probably right and they are not duplicates. Linking them would be good. I'll assign @oskgo to both, so he can look at them jointly.

  5. oskgo commented on Oct 14, 2025

    @oskgo
    Contributor

    They're not duplicates. The other is likely caused by mapping over an empty list of memory-formula pairs without providing an intended memory. This one is caused by not providing a memory when translating from expressions to formulas.

    I had similar issues before, but these must have slipped through.

    I'll have a closer look later today.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions