erts, kernel: Little-endian packet header lengths

[ptome]

Add support for little-endian packet header lengths in gen_tcp and erlang:decode_packet, for lengths of two, three and four.

Add support for packet header lengths of three, both LE and BE.

The packet option can now be a tuple {Len, Endianness}, where Len can be 2, 3 or 4 and Endianness can be big or little. If endianness is big the behavior is the same as the old options.

Requested in issue #7265.

Other notes:

• the tuple option was suggested in the issue, I added big for symmetry
• for the inet backend I used the enum meta-type and had to change the type_val_1 clauses for it to work; there could be a better way, the option now has different possible types
• the code I added in gen_tcp_socket:send could be simplified since at that stage the option is already validated?
• the tests for the tuple with big-endian option were commented out to reduce the total test time, but maybe they should be enabled; I tested locally, and they passed, I commented out after the fact.

[github-actions]

CT Test Results

    4 files    201 suites   1h 50m 25s ⏱️
3 119 tests 2 803 ✅ 306 💤 10 ❌
4 100 runs  3 678 ✅ 385 💤 37 ❌

For more details on these failures, see this check.

Results for commit ab146d6.

♻️ This comment has been updated with latest results.

To speed up review, make sure that you have read Contributing to Erlang/OTP and that all checks pass.

See the TESTING and DEVELOPMENT HowTo guides for details about how to run test locally.

Artifacts

• Complete CT logs (Download Logs)
• HTML Documentation (Download HTML Docs)
• Windows Installer

// Erlang/OTP Github Action Bot

[IngelaAndin]

This is something that should be discussed by OTP - technical board. I will put it on the schedule but as it is vacation period I think we will alas not have any conclusion before autumn, thankyou for your patience.

[ptome]

This is something that should be discussed by OTP - technical board. I will put it on the schedule but as it is vacation period I think we will alas not have any conclusion before autumn, thankyou for your patience.

I understand. I will wait for future updates, thank you.

[garazdawi]

My oppinion on this is that, as we are transitioning away from the inet driver, we should add this functionality in the socket backend via allowing a fun for packet. Then everyone can implement their own favourite packet mode without it having to be part of the stdlib.

[ptome]

Note: the failed checks are due to the missing preloaded bootstrap, as per project rules.

I think taking a fun as parameter is very interesting, but I'm not sure how it would work with erlang:decode_packet.

I will wait for further updates, thank you.