Move to fieldalignment linter from deprecated maligned linter

[dfarrell07]

What would you like to be added:

Move from the deprecated maligned linter to the new fieldalignment linter.

Why is this needed:

Upgrading to golangci-lint 1.39.0 in submariner-io/shipyard#519:

Deprecated linters:

• maligned, replaced by fieldalignment

[dfarrell07]

This is also reported as a warning in Go linting:

The linter 'maligned' is deprecated (since v1.38.0) due to: The repository of the linter has been archived by the owner. Replaced by govet 'fieldalignment'."

[skitt]

The switch itself is implemented in #1306, however it shows that fieldalignment flags more issues than maligned: it checks struct size, and “pointer size”, i.e. how much of the struct needs to be scanned for GC. This is really a minor optimisation (see golang/go#44877 for details) but it can’t be disabled.

Further, golangci-lint can use maligned’s suggestions, at least to present them to the user; fieldalignment is also capable of producing suggestions, but golangci-lint can’t use them yet (see golangci/golangci-lint#1779). Until that happens, switching to fieldalignment seems counter-productive to me: it is liable to find issues in structs without being able to explain how to fix them, and it does find lots of issues related to pointer coverage which are hard to fix.

[dfarrell07]

Okay, thanks for digging in and summarizing so clearly @skitt.

Can we add golangci-lint ignores for this linter matching text "pointer size" (or similar) to turn fieldalignment more into a drop-in replacement for maligned?

Further, golangci-lint can use maligned’s suggestions, at least to present them to the user; fieldalignment is also capable of producing suggestions, but golangci-lint can’t use them yet

I don't see either fieldalignment or maligned mentioned in the supported linters there, but I do think I remember maligned's suggestions.

So maybe the plan could be something like:

1. Wait until fieldalignment and maligned have suggestion parity.
2. Use ignore rules for "pointer size" if possible, otherwise wait on ability to disable that check, and then phase in the struct scanning optimizations if we want later.

[skitt]

Can we add golangci-lint ignores for this linter matching text "pointer size" (or similar) to turn fieldalignment more into a drop-in replacement for maligned?

Ah yes, I’d forgotten we could do that! I’ve added the relevant configuration in the updated PR. If it passes everything here I’ll mark it as ready for review.