`sort-prop-types` throws an error with a shape defined by a `const`

[leggechr]

The following code snippet:

const accountInfoType = {
  name: PropTypes.string,
  address: PropTypes.string,
  taxid: PropTypes.string,
};

AccountDetails.propTypes = {
  accountInfo: PropTypes.shape(accountInfoType),
}

Throws the following error:

Cannot read property 'reduce' of undefined
TypeError: Cannot read property 'reduce' of undefined
    at checkSorted ([localPath]/node_modules/eslint-plugin-react/lib/rules/sort-prop-types.js:80:19)
    at CallExpression ([localPath]/node_modules/eslint-plugin-react/lib/rules/sort-prop-types.js:166:9)
    at listeners.(anonymous function).forEach.listener ([localPath]/node_modules/eslint/lib/util/safe-emitter.js:47:58)
    at Array.forEach (native)
    at Object.emit ([localPath]/node_modules/eslint/lib/util/safe-emitter.js:47:38)
    at NodeEventGenerator.applySelector ([localPath]/node_modules/eslint/lib/util/node-event-generator.js:251:26)
    at NodeEventGenerator.applySelectors ([localPath]/node_modules/eslint/lib/util/node-event-generator.js:280:22)
    at NodeEventGenerator.enterNode ([localPath]/node_modules/eslint/lib/util/node-event-generator.js:294:14)
    at CodePathAnalyzer.enterNode ([localPath]/node_modules/eslint/lib/code-path-analysis/code-path-analyzer.js:608:23)
    at Traverser.enter [as _enter] ([localPath/node_modules/eslint/lib/linter.js:1006:32)

Declaring the propTypes in either of the following ways is a workaround so that it doesn't crash:

AccountDetails.propTypes = {
  accountInfo: PropTypes.shape({
    name: PropTypes.string,
    address: PropTypes.string,
    taxid: PropTypes.string,
  }),
}

const accountInfoType = PropTypes.shape({
  name: PropTypes.string,
  address: PropTypes.string,
  taxid: PropTypes.string,
});

AccountDetails.propTypes = {
  accountInfo: accountInfoType,
}

[ljharb]

When the value passed into shape is defined in the same file, it would be ideal to resolve it. Separately, if the value is defined elsewhere, we should ignore it instead of crashing.

[alexzherdev]

I've looked into this, and it seems like the error had been dealt with in #1669. That PR doesn't have any resolution logic though.