Test BEAM files without type information

[bjorng]

Normally, all BEAM files created by OTP 25 and later have a "Type" chunk that contains type information. beam_lib:strip/1 will not discard the "Type" chunk, but build/release scripts that do their own custom stripping could accidentally delete the chunk.

Make sure to test that loading and executing BEAM files without type information works. Since there was an overrun-heap-and-stack bug (reported in #7292, fixed in #7581) when using the bit syntax, the bit syntax test suites seems to be appropriate to clone to new BEAM files without types.

[github-actions]

CT Test Results

       4 files     425 suites   57m 52s ⏱️
2 166 tests 2 115 ✔️ 51 💤 0 ❌
6 877 runs  6 806 ✔️ 71 💤 0 ❌

Results for commit 3d04dbc.

♻️ 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

[RobinMorisset]

Is there an easy way to do such stripping of the Type chunk? I'd like to get it added as an option to erlfuzz.

[bjorng]

When this PR has been merged you can do erlc +strip_types some_file.erl.

If you want to do it yourself, the basic procedure is the following:

    {ok,_Module,Chunks0} = beam_lib:all_chunks(Beam0),
    Chunks = [{Tag,Contents} || {Tag,Contents} <- Chunks0,
                                Tag =/= "Type"],
    {ok,Beam} = beam_lib:build_module(Chunks)

where Beam0 can either be the name of a BEAM file or the contents of a BEAM file as a binary. Beam is always a binary containg the stripped BEAM module.

[bjorng]

In our daily builds, there were failed test cases in a few test runs (in 4 out of more than 50). The test cases failed because of uninitialised memory. The bug was hidden in most of the test runs because the freshly allocated memory happened to be zero. The bug is fixed in #7616.