fix: single sequencer

[edobry]

This PR fixes a logic bug when provisioning network participants where all participants were mistakenly configured as sequencers.

Follow-up to this PR.

Context

[SozinM]

Hi! @edobry
Does this line contains any side effects?

if sequencer_enabled:
            sequencer_enabled = False

Because it seems strange to change the variable that isn't used after (but i'm not that familiar with starlark)

Also the reason i used insert(0) is because later we use things like all_el_contexts[0] to choose EL to which batcher would connect. So we must ensure that main EL instance (probably sequencer) is in first position in this list.
It causes side effects when code is changed by someone who doesn't know this quirk (that's why i put this comment in place)

[sigma]

Hi! @edobry Does this line contains any side effects?

if sequencer_enabled:
            sequencer_enabled = False

Because it seems strange to change the variable that isn't used after (but i'm not that familiar with starlark)

Also the reason i used insert(0) is because later we use things like all_el_contexts[0] to choose EL to which batcher would connect. So we must ensure that main EL instance (probably sequencer) is in first position in this list. It causes side effects when code is changed by someone who doesn't know this quirk (that's why i put this comment in place)

I'm a bit confused. This code is in the participants loop, so the way I'm reading it:

• flipping sequencer_enabled results in sequencer_enabled being False for all iterations but the 1st one (effectively making the 1st node the sequencer, which I think is aligned with what you say)
• doing insert(0, ...) repeatedly effectively makes the last participant be the 1st element of the resulting list. Which seems to be the opposite of what that comment stated.

I must be missing something.

As a meta-note, maybe we should refactor this code to return something like

struct(
  sequencer=...
  all_contexts=...
)

So that we don't have to rely on an index-based convention elsewhere. WDYT?

[SozinM]

Ah, got it!
I didn't saw the loop and assumed it was the end of function, now it become much clearer, thank you :)