[3.6] Pr for issue 10434

[jbsgh]

Description

Improve usage of rand() for certain LCG-implementations.

PR checklist

I did not build it nor run any tests because our target differs from the officially supported targets in many aspects. I've just backported our changes of the source code.

I've used Tag v3.6.4 as base.

[jbsgh]

I've branched tag v3.6.4 for preparing the PR

[mpg]

Looks good to me, thanks!

As discussed in the issue, I don't think that's what we want in the long term, but in the meantime it improves the situation on some platforms while not making anything worse, so it passes the bar for inclusion.

[mpg]

@jbsgh As indicated in our contributing guidelines we require a Signed-off-by: line in every commit in order to indicate compliance & agreement with our contribution requirements. Can you amend your commit to include that in the message? I guess git commit --amend --signoff should work, then you'll need to force-push. Thanks!

[mpg]

(Once that's done, please remind me to kick the CI.)

[jbsgh]

Done. @mpg: Reminder!

[mpg]

Thanks for the reminder! I've just started the jobs, results should be back in a few hours.

[mpg]

The CI came back with a number of failures. You can ignore all failures in components with armcc in their name, as they are due to some temporary licensing issues, but the other errors will need fixing.

One error that was reported was:

[2025-11-07T10:29:01.134Z] test/benchmark.c:458:9: error: unused variable 'rnd' [-Werror,-Wunused-variable]
[2025-11-07T10:29:01.134Z]   458 |     int rnd;
[2025-11-07T10:29:01.134Z]       |         ^~~
[2025-11-07T10:29:01.654Z] 1 error generated.

but there might be others. I think I'd recommend fixing this one first, then running the CI again and we'll see what (if anything) is left.

[mpg]

Thanks for the update! However our DCO checker points out the second commit is missing the Signed-Off-By line.

Also, we're having trouble with our CI at the moment. I'll kick it once it's stable again, sorry for the delay.

[mpg]

It looks like something went wrong when fixing the commit message, now the history still has the original non-signed-off commit, plus the same with a sign-off, and a merge commit. Could you fix the history and force-push? Then I'll kick the CI (we got it fixed in the meantime).

[jbsgh]

What do you want me to do in detail? Please specify the requested git commands to fix your CI issues on the PR.

[mpg]

What do you want me to do in detail?

We require every single commit to include a signed-off-by line in its message. Right now your PR has 4 commits, two of which have this line, and two of which don't.

Furthermore, the current history is confusing because there are two commits with the same summary and diff, and also a merge commit for no good reason.

Please specify the requested git commands

Well, off the top of my head, the following should do the trick (untested, use at your own risk):

git checkout pr-for-issue-10434
git reset --hard c2d0580613b9283ad45fabdbb0b1ff63794068ab
git commit --amend --sign-off
git push --force-with-lease <name_of_your_remote> pr-for-issue-10434

After doing that, the DCO line in the status reported on github should be green.

Once we're reached that stage, I'll kick the actual CI.