check if what4 adapter is present before setting timeout

[danmatichuk]

fixes #1807

[RyanGlScott]

Instead of having to enumerate all of the solvers and their timeout options, I wonder if we should do the following:

1. Add a new field to AnOnlineAdapter (of type ConfigOption BaseIntegerType) to represent the timeout option for a particular solver.
2. In setTimeout, match on cfg and apply the timeout option for each solver that is contained in the W4ProverConfig.

This approach has the advantage that we wouldn't need to remember to keep the implementation of this function in sync whenever Cryptol adds support for a new SMT solver. (In fact, I suspect that this function is already out of sync, since this function doesn't mention the recently-added Bitwuzla solver anywhere.)

[danmatichuk]

This makes sense to me, although I think we'll need an auxiliary config option for yices to account for the second/millisecond discrepancy vs. the rest of the solvers.

[RyanGlScott]

Are you able to reproduce this test failure from CI locally? If so, do you know what might be causing it?

      issue_1807.icry: [Failed]
6c6,8
< Solver returned UNKNOWN
---
> 
> What4 exception:
> fd:18: hFlush: resource vanished (Broken pipe)

(I do not use macOS, so I am unable to reproduce it myself.)

[danmatichuk]

I've seen this before and it's definitely an issue with What4 itself. It's highly nondeterministic from what I can tell.

I'll see if I can reproduce it on an up to date What4 build and raise a corresponding issue there. I believe it's limited to the cvc solvers, so we can remove those from the test for now.

[danmatichuk]

Currently failing as a result of GaloisInc/what4-solvers#58. Should work with an updated build of yices.

[RyanGlScott]

I've published a new what4-solvers snapshot (snapshot-20250227) that includes a fix for GaloisInc/what4-solvers#58. I've also pushed a commit that makes use of it—let's see what happens.

[RyanGlScott]

Hooray! That appears to have fixed it!

As always, please make sure to mention this bug fix in the changelog. Otherwise, LGTM.