Pattern synonym in record update throws syntax error

Summary

Using a pattern synonym in a record update results in a syntax error. However, wrapping the pattern synonym in parentheses works. This is surely unintended behavior.

Steps to reproduce

{-# LANGUAGE PatternSynonyms #-}

data R = R { x :: () }

pattern P :: R
pattern P = R { x = () }

error = P { x = () }  -- error: Constructor ‘P’ does not have field ‘x’
works = (P) { x = () }  -- workaround

Expected behavior

Being able to use pattern synonyms directly in record updates without needing parentheses.

Environment

• GHC version used: any

Discussion on libera.chat:#haskell

23:02:06  Irssi: Join to #haskell was synced in 69 secs
23:06:47  <yin> this is fun!: i have a bidirectional pattern synonym and try to use record update syntax like `MyPattern { x = ... }` i can get the error `constructor MyPattern does not have field 'x'`
23:07:10  <yin> the solution: `(MyPattern) { x = ... }`
23:08:24  <yin> this feels kind of lispy
23:10:51  <yin> also feels like PatternSynonyms could be patched to allow the former
23:32:35  <c_wraith> that definitely sounds like a feature it should have.
23:32:54  <c_wraith> If it allows you to create a record pattern, it should use the same syntax as a record pattern...
23:33:05  <monochrom> To me it feels like kind of CPP.  "#define add(x,y) ((x)+(y))" :)
23:33:49  <monochrom> PPP = pattern pre-processor >:)
23:36:51  <geekosaur> record update syntax is a bit weird
23:37:38  <c_wraith> like at the parser level?
23:37:41  <monochrom> It's probably a simple oversight in the parser.
23:39:37  <geekosaur> yeh, it's probably missing a case for patsyns so it makes you fall into the parenthesized expression case
...
23:47:13  <geekosaur> anyway it sounds like someone should file a bug on gitlab