Are there any compiler warnings that could have prevented this?

Isn't there a warning saying a do-block expression is discarding a result? I'd have expected that to fire here.

Isn't there a warning saying a do-block expression is discarding a result? I'd have expected that to fire here.

Yeah, same, but it didn't. No idea why not.

I thought we are already using -Wunused-do-bind, but maybe we aren't, I'll give it a try.

I thought we are already using -Wunused-do-bind, but maybe we aren't, I'll give it a try.

In this pr? or could we merge it?

This is interesting. I don't know what is going on.

EDIT: this is what the ghc invocation looks like:

/home/hugin/.ghcup/bin/ghc --make -fbuilding-cabal-package -O -static -dynamic-too -dynosuf dyn_o -dynhisuf dyn_hi ... -XHaskell2010 -XApplicativeDo -XBangPatterns -XDeriveFunctor -XDeriveGeneric -XDeriveFoldable -XDeriveTraversable -XFlexibleContexts -XGeneralizedNewtypeDeriving -XLambdaCase -XNamedFieldPuns -XOverloadedStrings -XRecordWildCards -XScopedTypeVariables -XStandaloneDeriving -XTupleSections -XTypeApplications -XViewPatterns -XDataKinds -XTypeOperators -XKindSignatures Control.Concurrent.Strict ...  Development.IDE.Session ... -Wunused-do-bind -hide-all-packages -haddock -Wunused-do-bind

contains -Wunused-do-bind twice, as I supply it via cabal.project and ghcide.cabal.

Let's look into some more details before merging, imo.

Very interesting:

I don't understand, copying that code to another project highlights correctly both occurrences. Even deleting everything except these two functions doesnt work for me, i.e. still shows only one warning.

I found it: it's ApplicativeDo, which is enabled in ghcide.cabal. Please disable it @fendor since it doesn't make any difference with hls-graph anymore.