Strange block.timestamp / warp interactions with via-ir

[marktoda]

Component

Forge

Have you ensured that all of these are up to date?

• Foundry
• Foundryup

What version of Foundry are you on?

forge 0.2.0 (f6361e6 2022-12-01T15:11:38.140433224Z)

What command(s) is the bug in?

forge test

Operating System

Linux

Describe the bug

Block.timestamp and warp seem to have very strange behavior under via-ir. Minimal repro:

foundry.toml:

• via_ir = true

    function testWarp() public {
        console2.log("timestamp first", block.timestamp);
        vm.warp(block.timestamp + 100);
        console2.log("timestamp, timestamp + 100", block.timestamp, block.timestamp + 100);
    }

Expected output:

timestamp first: 1
timestamp, timestamp + 100: 101 201

Actual output:

timestamp first: 1
timestamp, timestamp + 100: 101 101

Super strange - I have no idea why the addition would not apply to block.timestamp

Another repro showing that stack variables get clobbered

    function testWarp() public {
        uint256 timestamp = block.timestamp;
        console2.log("timestamp before", timestamp);
        vm.warp(timestamp + 50);
        console2.log("timestamp after", timestamp);
    }

Expected output:

timestamp before 1
timestamp after 1

Actual output:

timestamp before 1
timestamp after 51

[marktoda]

oh also note that setting via_ir = false fixes the issue. Removing the warp also fixes the issue

[marktoda]

#1373

seems compiler is seeing my block.timestamp + 100 pre-warp and autoresolving my block.timestamp + 100 post-warp to the same value

[marktoda]

Can close i suppose but this is super strange behavior :(

[marktoda]

Any way we could remove whatever optimizerStep that does this clobbering for test/ contracts?

[PaulRBerg]

@marktoda did you find any workaround, after all?

I just enabled --via-ir and bumped into the same problem as you did here .. it looks like there is now an open issue in the Solidity repo (ethereum/solidity#12963) to track this strange behavior of Solidity, but I am not quite sure what I should do in the meantime? I would prefer not to disable --via-ir.

Did you, by chance, override the optimizerSteps in your Foundry config? I noticed that if I set it to either `` or dhfoDgvulfnTUtnIf, I get the strange behavior reported by you. But if I remove this configuration completely, the `warp` cheatcode works as expected.

[marktoda]

Hey @PaulRBerg !

Honestly my solution was just to redesign the test suite to get around this issue. In our case, changing some constants around to re-evaluate block.timestamp in different expressions allowed the value to change properly, i.e.

uint256 startTimestamp = block.timestamp;
... startTimestamp + 10 // 11
uint256 newTimestamp = startTimestamp + 100); // 101
warp(newTimestamp)
... startTimestamp + 50  // 51

I did tinker around with optimizerSteps, but couldn't isolate the correct step easily and gave up to do the hacky way out :D

[PaulRBerg]

Thanks for the tips, @marktoda. I will probably end up doing the same, i.e. introduce a few new constants to go around this whacky behavior of warp.