Fix the Python RPC tests on Big Endian architectures

[peternewman]

• New placeholder for NEWS
• Fix a typo in NEWS
• Fix the OS X builds on Travis
• Add a missing semicolon to the build for if we don't have CLOCK_MONOTONIC

The above has all been moved into #1703

This is now mostly:

• Fix the Python RPC tests on Big Endian architectures

[peternewman]

We probably want to cherry-pick this commit into here too:
5286a69

Given the issues in Homebrew/homebrew-core#65415

Now also in #1703

[peternewman]

@yoe 91cb697 should fix those test on Big Endian architecture, would you mind testing it somehow as I've only dummied it up by reversing stuff on an LE machine. @brucelowekamp care to take a look at it for any Python howlers I've introduced?

[peternewman]

Technically we could just run it all of the time, as it wouldn't change the order on an LE machine, but that seemed a bit unnecessary

[brucelowekamp]

@yoe 91cb697 should fix those test on Big Endian architecture, would you mind testing it somehow as I've only dummied it up by reversing stuff on an LE machine. @brucelowekamp care to take a look at it for any Python howlers I've introduced?

I don't know of any python concerns, but I do have to ask if this isn't a protocol error? It's sent out on the socket, so if you were to use the API across two endian machines, it wouldn't work? Looks like both the Python and C++ do the same thing, but still....

Bruce

[peternewman]

I don't know of any python concerns, but I do have to ask if this isn't a protocol error? It's sent out on the socket, so if you were to use the API across two endian machines, it wouldn't work? Looks like both the Python and C++ do the same thing, but still....

I'd be tempted to agree with you @brucelowekamp and I did consider fixing it. However due to the limitations mentioned here:
http://docs.openlighting.org/ola/doc/latest/rpc_system.html#rpc_OLA_RPC

There won't/shouldn't be any inter-machine comms and hence any multi-endian issues. Whereas if we do fix it, then if someone runs an unfixed client against a fixed server, it won't work. Although given this is only on BE machines, I guess it's a fairly low risk. It will definitely be a breaking change though, so I'm not sure I'm personally convinced of the benefit apart from hypothetically. I guess we should at least flag the lines of code with a TODO and possibly consider making the change for 0.11.0 or similar.

Unfortunately @yoe found it didn't entirely work:

(10:52:23) wouter: PeterNewman4: your patches don't seem to resolve the issue
(10:52:31) wouter: that is, it kind of works, but things seem to hang
(10:52:58) wouter: and I also get a line saying "WARNING:root:Protocol mismatch: 3 != 1" when I run python/ola/RDMTest.py

Having had a better look, that code is in our codebase:

ola/python/ola/rpc/StreamRpcChannel.py

Lines 246 to 250 in d101097

	if version != self.PROTOCOL_VERSION:
	ola_logger.warning('Protocol mismatch: %d != %d', version,
	self.PROTOCOL_VERSION)
	self._skip_message = True
	self._expected_size = size

Given it's the RDM ones, presumably it's of course going to be because they have responses too and I haven't mangled those responses on the way back out, so it can't read them on BE architectures/thinks they're corrupt!

[peternewman]

Okay @yoe this definitely cracks it. Here's a broken build on s390x (thanks to Travis):
https://travis-ci.org/github/OpenLightingProject/ola/jobs/748167069#L4381-L4475

And here's one with that fix applied on top (still on the same architecture):
https://travis-ci.org/github/peternewman/ola/builds/748170563

[yoe]

Awesome. Didn't know Travis supports s390x too these days. Very cool! Am I correct in stating that I just need to cherry pick ead457c, eb1f26a, and c402fcb? If not, what else would I need? Peter Newman <notifications@github.com> schreef op 7 december 2020 22:29:08 GMT+02:00:

…

Okay @yoe this definitely cracks it. Here's a broken build on s390x (thanks to Travis): https://travis-ci.org/github/OpenLightingProject/ola/jobs/748167069#L4381-L4475 And here's one with that fix applied on top (still on the same architecture): https://travis-ci.org/github/peternewman/ola/builds/748170563 -- You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub: #1687 (comment)

-- Verstuurd vanaf mijn Android apparaat met K-9 Mail. Excuseer mijn beknoptheid.