client/core: resolve xcWallet.peerCount race (parent part)

[norwnd]

Race detector found another race I missed (this one is a bit more involved, having to do with reading parent wallet state without acquiring parent mutex) when addressing similar one.

WARNING: DATA RACE
Write at 0x00c00047a300 by goroutine 5580:
  decred.org/dcrdex/client/core.(*xcWallet).Lock()
      /Users/norwnd/dcrdex/client/core/wallet.go:193 +0x1f5
  decred.org/dcrdex/client/core.(*xcWallet).Lock()
      /Users/norwnd/dcrdex/client/core/wallet.go:181 +0x184
  decred.org/dcrdex/client/core.(*Core).Logout()
      /Users/norwnd/dcrdex/client/core/core.go:4735 +0x352
  main.promptShutdown.func1()
      /Users/norwnd/dcrdex/client/cmd/dexc/main.go:241 +0x3d

Previous read at 0x00c00047a300 by goroutine 61:
  decred.org/dcrdex/client/core.(*xcWallet).state()
      /Users/norwnd/dcrdex/client/core/wallet.go:271 +0x217
  decred.org/dcrdex/client/core.(*Core).assetMap()
      /Users/norwnd/dcrdex/client/core/core.go:2258 +0x727
  decred.org/dcrdex/client/core.(*Core).SupportedAssets()
      /Users/norwnd/dcrdex/client/core/core.go:2209 +0x189
  decred.org/dcrdex/client/core.(*Core).refreshFiatRates()
      /Users/norwnd/dcrdex/client/core/core.go:9744 +0x17c
  decred.org/dcrdex/client/core.(*Core).fetchFiatExchangeRates.func1()
      /Users/norwnd/dcrdex/client/core/core.go:9725 +0x14e

[norwnd]

Whoops, forgot to add the main change itself, here it is: 835df7c,

so all 3 commits can be squashed before merging to master, can you do this on merge or do I need to force-push for this myself (just so I know in future) ?

[chappjc]

so all 3 commits can be squashed before merging to master, can you do this on merge or do I need to force-push for this myself (just so I know in future) ?

It's fine to squash in this case (just put up the pr and did change X followed by undo change X, so just squash). I like to have separate commits when either:

• they are conceptually separate and aid review, and I intend to have them merged separately
• they are review updates, which I generally squash after the reviewer has seen them.

[chappjc]

LGTM.

I need to get back to running with the race detector more often.

In addition to finding data races, it also has the side effect of making you acutely aware of operations that can get slow, especially on mainnet or testnet or large wallets. Just playing around on simnet hides many real life bottlenecks, so I'm glad you spend time working with testnet.