Does the test only support little endianess?

[rperlas]

When I attempt an NPM on big endian machines, I run into two failing test cases.

One error is on the command: "node --expose-gc nannew-test.js".

Another error is:

not ok 69 ../cpp/nannew.cpp:323: stringMatches( New(widestring, 4).ToLocalChecked(), "Hell")
---
file: events.js
line: 72
column: 20
stack:
- |
getCaller (/home/nodebot/node_modules/nan/node_modules/tap/lib/tap-assert.js:439:17)
- |
Function.assert (/home/nodebot/node_modules/nan/node_modules/tap/lib/tap-assert.js:21:16)
- |
Test._testAssert as ok
- |
emitNone (events.js:72:20)
- |
Test.emit (events.js:166:7)
- |
Test.emit (/home/nodebot/node_modules/nan/node_modules/tap/lib/tap-test.js:104:8)
- |
GlobalHarness.Harness.process (/home/nodebot/node_modules/nan/node_modules/tap/lib/tap-harness.js:87:13)
- |
doNTCallback0 (node.js:417:9)
- |
process._tickCallback (node.js:346:13)
- |
emitNone (events.js:72:20)

10:54:20 AM: not ok 95 test/js/nannew-test.js

I am wondering if this is because big endianess is not supported?

[kkoopa]

Don't know. See #481. We can't test on big endian, so we don't really bother. The library might still work even if the tests don't.

[bnoordhuis]

The node.js CI has BE machines in its matrix. We could move over if needed, details TBD.

[joransiu]

Hmm, this test failure looks a bit different than #481. #481 specifically sets the encoding as UCS2 (aka UTF16le). In this case, the test is simply creating a uint16_t string, with the chars encoded in little endian format.

  const uint16_t *widestring = reinterpret_cast<const uint16_t *>("H\0e\0l\0l\0o\0");
  t.ok(_( stringMatches( New(widestring, 4).ToLocalChecked(), "Hell")));`

https://github.com/nodejs/nan/blob/master/test/cpp/nannew.cpp#L322-L323