@ShahanaFarooqui has investigated the CI failures... Christian and Rusty are working on resolving some CI problems right now.
Once their fixes are applied, we'll get you to rebase and take a closer look if it still fails.

Shahana also noticed you forgot to add a Changelog entry in the commit.... here is the guide. You'll need to fix this: a very important habit! Not adding a Changelog entry can also fail CI (when it's working as expected).

Well done and congratulations on your first PR for CLN!

@dovgopoly Thanks for the PR! I updated your commit message to include a changelog entry. Normally, the CI would fail with a clear error like:

::error::'Changelog' entry is missing in all commits, and 'Changelog-None' not specified in the PR description

However, for some strange reason, the CI didn’t even start the pre-build this time. Adding the changelog message allowed the CI to proceed, and we can investigate further if this issue happens again.

By the way, feel free to edit the commit message as you see fit, I just added it to test the CI flow. :).

Thanks! Great find!

Since this is your first PR I'm going to nitpick!

1. Please write a quick test (perhaps in tests/test_misc.py) which triggers this: you can have it call out to lightning-cli exactly as you wrote here.
2. Please include the crash backtrace in the commit msg: this is useful fodder for anyone looking at a similar crash!
3. Changelog should probably say "Changelog-Fixed: JSON-RPC: fix crash when invalid filter parameter passed in JSON request" since we usually try to start with sub-categories, and CHANGELOG is for users, not (internal) developers.

1. This segfault is only happened when the --developer mode is off. So it's hard to trigger it in tests. The only approach I found:

def test_filter_with_invalid_json_bug(node_factory):
    l1 = node_factory.get_node(start=False)
    l1.daemon.early_opts = []
    l1.daemon.opts = {k: v for k, v in l1.daemon.opts.items() if not k.startswith('dev')}
    l1.start()

    ret = subprocess.run(['cli/lightning-cli',
                          '--network={}'.format(TEST_NETWORK),
                          '--lightning-dir={}'
                          .format(l1.daemon.lightning_dir),
                          '-l', '1',
                          '-k',
                          'wait',
                          'subsystem=invoices',
                          'indexname=created',
                          'nextvalue=0'])

2-3. Fixed.

And here's the blog post I made about it, with video!

https://rusty.ozlabs.org/2025/12/12/CLN-Neatening-A-Bugfix-PR.html

Thanks! It was an awesome blog post :)
I was unsure whether we need to add a test which turns off the dev mode. After your video, I understood that it's not bad. Added some assertions there as well.
I've also added a backtrace to the commit message.
cc @rustyrussell