<leonardoalt> btw POPL22 is live with streams:
https://popl22.sigplan.org/program/program-POPL-2022/?date=Wed%2019%20Jan%202022&past=Show%20upcoming%20events%20only
https://popl22.sigplan.org/program/program-POPL-2022/?date=Wed%2019%20Jan%202022&past=Show%20upcoming%20events%20only
<ehildenb> Well gosh darn, it worked! It shows the second loop (the actual for-loop I wrote) 3 times because it took two iterations of the first loop (the abi decoding) to find out that it was indeed a loop. So the first copy of the last loop is "if there were no iterations of the first loop", the second one is "if there was one iteration of the first loop", and the last one is "if there were N>1 iterations of the first loop".
https://cdn.discordapp.com/attachments/798403407076917269/933507949320601690/Screenshot_from_2022-01-19_16-45-02.png
https://cdn.discordapp.com/attachments/798403407076917269/933507949320601690/Screenshot_from_2022-01-19_16-45-02.png
<ehildenb> So that's all redundant work, but I'm pretty stoked this one worked! Did not expect that.
<Juan C> Hi! I'm creating a Docker image to use the SMTChecker with Eldarica so that more people can try this configuration without having to go through the modification steps. Currently I'm downloading the
run_small.js script from github and modifying it while building the image. I wanted to ask you if it was okay to include this file (via download or just the plain file) in the image repo
<leonardoalt> imo the main annoying part is that we still need the custom solc-js branch to access the solver callback and eldarica properly. Hopefully we'll merge the PR today though, so you only need solc-js master branch
<leonardoalt> also currently the eldarica set up does not output counterexamples/invariants, only the result. Those 2 are under development, I have a branch with some code for that but not ready yet
<leonardoalt> + the js script that I pasted (gist), modified to filtering
smtsolver.availableSolvers to eldarica (or whichever solver you want) + modified settings
<leonardoalt> + a recent soljson.js release
<Juan C> Sure! I'll be sure to ping you when it's ready ๐
<leonardoalt> @ehildenb @Juan C @Kademlia and others:
We're inviting tooling authors first to register:
https://forms.gle/3hKAxYuQSVGeoX7s5
We're inviting tooling authors first to register:
https://forms.gle/3hKAxYuQSVGeoX7s5
<ehildenb> Either way I'll make my way to Amsterdam, and I'm sure we'll all run into each other!
<ehildenb> Also, if anyone here is going to Eth Denver, I live 1.5 hours from there, would be down to meet up!
<leonardoalt> @ehildenb @anvacaru sorry for the delay, but here it is:
https://output.circle-artifacts.com/output/job/f0d73a61-d44a-4070-84a1-d1229293600f/artifacts/0/soljson.js
you can try to run Eldarica with this soljson, it has some support to counterexamples and invariants from Eldarica.
Please let me know if it works!
https://output.circle-artifacts.com/output/job/f0d73a61-d44a-4070-84a1-d1229293600f/artifacts/0/soljson.js
you can try to run Eldarica with this soljson, it has some support to counterexamples and invariants from Eldarica.
Please let me know if it works!
<Kademlia> putting together a collection of examples of programming proofs, proof procedures, etc. Dropping here in case it is interesting to anyone https://github.com/WilfredTA/proof-programming
<d-xo> which seems to suggest that
>>Int is translated to shrInt in smt, but I can't find a definition of shrInt anywhere?
<ehildenb> For the most part, we try do deal with any bitwise operations via algebraic reasoning (rewrites over expressions), rather than some bit-blasting approach or something.
<ehildenb> And I think it's the correct way to do it.
<ehildenb> Bit-blasting I think makes very little sense for software verification honestly. Most software is not doing bit-level manipulations.
<ehildenb> And EVM bytecode that comes from the Solidity compiler (and likely other compilers) have very regular structure.
<ehildenb> So bit-level operations on words are to project out chunks of bits for packed storage slots for the most part.
<ehildenb> Which can be represented algebraically much cleaner.
<ehildenb> For example, something like
maxUInt48 &Int (X >>Int 48) means "give me the second lowest 48 bits"
<ehildenb> And if that's what you want, then you can just have a rule like
rule maxUInt48 &Int (X >>Int 48) => projectSecondLowest48(X)
<ehildenb> Then axiomatize
projectSecondLowest48 directly.
<ehildenb> Or you can have more generic
project... functions which can be used.
<ehildenb> But the point is to take advantage of the structure that the compiler introduces. It's not like it will introduce arbitrary bitwise operation, but highly structured ones. bit-blasting gives up completing on understanding that structure.
<ehildenb> Here, for example, is the algerbraic presentation of the needed packing/unpacking lemmas for multi-collateral dai: https://github.com/runtimeverification/evm-semantics/blob/master/tests/specs/mcd/word-pack.k
<anvacaru> @leonardoalt I've changed the Dockerfile-eldarica to use your suggested soljson file. When an assertion violation happens I get a large output which starts with
{
auxiliaryInputRequested: {
smtlib2queries: {
'0x5e7df8f9c2da419124ac9c1918697f4d25bea8f5918512e44855c3ee108eb46f': '(set-logic HORN)\n' +
'\n' +
'(declare-datatypes ((|state_type| 0)) (((|state_type| (|balances| (Array Int Int))))))\n' +.
I did not get any counter examples. Instead, I get a message at the end saying that there were no Horn solvers available.
{
component: 'general',
errorCode: '3996',
formattedMessage: 'Warning: CHC analysis was not possible. No Horn solver was available. None of the installed solvers was enabled.\n' +
'\n',
message: 'CHC analysis was not possible. No Horn solver was available. None of the installed solvers was enabled.',
severity: 'warning',
type: 'Warning'
}
],
<leonardoalt> which is very weird if the script didn't change