Fix raw-filter flags by binding flags to viper only if command is run

[glrf]

The raw-filter flags are broken since we bind them to viper twice with the same name, during initalization. Once for hubble observe and once for hubble observe flows. This means that only the flags for hubble observe flows work, as it is registered later and overwrites the binding for hubble observe

We fix this by only binding the flags of the command we actually run.

Fixes: #1201

[glrf]

I'm not entirely sure this is the best way to fix this. It's probably the shortest, but it doesn't remove the original footgun that we can easily bind two flagsets to the same viper config.
Maybe someone has a better idea?

[chancez]

This approach seems reasonable. I'm sure there's probably a "better" way to do it, but I suspect it involves bigger refactors.

I do think what we need however is a test for this. Could you add a new test case into https://github.com/cilium/hubble/blob/main/cmd/cli_test.go where you call hubble observe --allowlist ... --print-raw-filters and verify the output contains the expected "raw filters"? I think that's probably the simplest. We should have a test case for both hubble observe and hubble observe flows to make sure they're both getting the raw filters.

[glrf]

I do think what we need however is a test for this. Could you add a new test case into https://github.com/cilium/hubble/blob/main/cmd/cli_test.go where you call hubble observe --allowlist ... --print-raw-filters and verify the output contains the expected "raw filters"? I think that's probably the simplest. We should have a test case for both hubble observe and hubble observe flows to make sure they're both getting the raw filters.

Good point, I totally forgot about that. I added a test case for it. Had to do a very minor change to printing raw filters toactually use the output defined by the command, so that the test case actually works.

[kaworu]

Nice catch!