bug(doc): forge doc seg faults

[sandybradley]

Component

Forge

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

• Foundry
• Foundryup

What version of Foundry are you on?

1.4.0-nightly

What version of Foundryup are you on?

No response

What command(s) is the bug in?

forge lint, forge fmt --check, forge doc --build

Operating System

Linux

Describe the bug

3 bugs in last 2 weeks:

• forge lint: forge lint throws Error: solar run failed with latest nightly build #11668
• forge fmt --check: not using foundry.toml settings
• forge doc --build:

thread '<unknown>' has overflowed its stack
fatal runtime error: stack overflow, aborting

These checks have been running for months without issues. Now they are failing from some recent foundry updates.

[grandizzy]

@sandybradley thanks for reporting! indeed we rolled out a new formatter which could produce 2nd and 3rd points, is the repo you see the issues public so we investigate? The first one we provided a fix for, please retest and let us know should still be issues.

[grandizzy]

Also mind that nightly is subject of active development so such situations are expected, you may want to pin stable in CI and test release candidate builds when announced in order to avoid such. Thanks for helping us improving Foundry!

[sandybradley]

@sandybradley thanks for reporting! indeed we rolled out a new formatter which could produce 2nd and 3rd points, is the repo you see the issues public so we investigate? The first one we provided a fix for, please retest and let us know should still be issues.

Thank you. Sadly not public yet. Yes, I took your advice and disabled the linter for the first one. I will try using stable a release to fix the other 2.

[grandizzy]

@sandybradley Could you then provide the entire output of docs crash and forge fmt, can help figuring out the problem. For point 1, besides the workaround, nightly contains a fix so worth trying locally and let us know if it is fixed. Thanks!

[grandizzy]

I was actually able to repro the forge doc issue and will provide a fix for. Would be great if you can retest the forge lint one with latest nightly and provide some details on which settings is not used when forge fmt --check. Thank you!

[grandizzy]

backtraces from forge doc seg fault on https://github.com/Uniswap/contracts repo bt.txt with a simple Counter project seems there's an infinite loop trying to fmt / parse sources from stdin

foundry/crates/doc/src/builder.rs

Line 137 in fe10724

forge_fmt::format(&source, self.fmt.clone()).into_result().unwrap_or(source);

trace.txt

[sandybradley]

I was actually able to repro the forge doc issue and will provide a fix for. Would be great if you can retest the forge lint one with latest nightly and provide some details on which settings is not used when forge fmt --check. Thank you!

Confirming first that switching to stable release in the ci did fix the immediate issue. Second, forge lint seems to work now on latest nightly. I see a bunch of stuff added there which is nice. Third, forge fmt --check did not seem to read line_length in the foundry.toml setting or perhaps there is a different spacing default. In any case it gives different results than the last stable release. Forth, thanks for following up with forge doc.

[grandizzy]

thank you, mind to pass fmt settings you use? There could be similar case with #11933 where new fmt alligns on properties and you need params_first_multi instead params_first