this does not change any functionality, it removes a case where we
initially lie about which values is `others_num`

these types do not exist in V2, which can be verified either by the
reading the bandwidth docs here[0] or by reading the commit message of cd6e0df

[0]: https://dev.bandwidth.com/docs/messaging/webhooks/#inbound-message-webhooks

pry-byebug explicitly says that pry-stack_explorer is incompatible

* 'emit-resends-as-resends' of https://git.secluded.site/sgx-bwmsgsv2:
  Emit ResendIn event when replaying inbound messages

* 'incorporate-first-round-of-hotfixes' of https://git.secluded.site/sgx-bwmsgsv2:
  fix: make sure ARGV[0] is set for all tests
  fix: don't call `from=` on StanzaError
  fix: Message::Base.media_urls is always an array
  fix: correct for dedup in message event
  fix: webhook tests need to stub component's `write`
  fix: webhooks expect top-level 'to'
  fix: response_handler wants unescaped strings
  fix: don't stop reactor if test

Add a ResendIn message event type that records the original stream ID
and bandwidth ID without duplicating body or media fields. The webhook
handler emits this event instead of a normal In event when the
X-JMP-Resend header is present.

except we also unescape each individual tag part?

each test runs on its own reactor. stopping it means that `assert`s are
never called, which meant that gateway shutdown was counted as test
success. no bueno!

includes the same fix as better-feature-order

so I stop accidentally blowing it away

pending bigger move to newer Ruby version which supports Gems that don't
rely on old, broken APIs from the stdlib

* 'catch-errors-from-at-least-bandwidth' of https://git.secluded.site/sgx-bwmsgsv2:
  Adopt cmd error handling like sgx-jmp

Might be short codes for example

* 'emit-messages-to-event-stream' of https://git.secluded.site/sgx-bwmsgsv2:
  Separate message_id into bandwidth/stanza IDs, make timestamp optional
  Switch from ValueSemantics to normal OOP Ruby
  Use stream name in example
  Remove unused wait_for
  Pin ruby-bandwidth-iris to ~> 2.7
  Add Redis stream trim example
  Emit messages and statuses to Redis stream

- The first commit was broken since it didn't include the `with_stubs`
  helper

- The second one was included in a previous patchset but I'm including
  it as a standalone to get upstream tests 'fixed.' Apparnetly this test
  reliably passed without stubbing TnOptiions.get, but clearly it should
  be stubbed or mocked.

The following changes since commit 710dfae450ea52a53a46f92776451e9e3f1afa19:

  Merge branch '380-conditionally-show-command' of https://git.secluded.site/sgx-bwmsgsv2 (2025-10-22 09:49:51 -0500)

are available in the Git repository at:

  test_fixes

for you to fetch changes up to 23cca00d8274c7f8bf808d5f0af6993369c7a76d:

  fix `test_port_out_pin` (2026-01-31 10:17:03 -0500)

----------------------------------------------------------------
Phillip Davis (2):
      fix: handle 404 in tn_eligible_for_port_out_pin?
      fix `test_port_out_pin`

 lib/bandwidth_tn_options.rb       |  8 +++++++-
 test/test_bandwidth_tn_options.rb | 19 +++++++++++++++++++
 test/test_component.rb            | 34 ++++++++++++++++++++++++----------
 test/test_helper.rb               | 11 +++++++++++
 4 files changed, 61 insertions(+), 11 deletions(-)
 create mode 100644 test/test_bandwidth_tn_options.rb

So it doesn't fall over on errors and we have to rely on a supervisor to
bring it back up a couple minutes after. This also adds Sentry error
reporting, only tested with the mock, not with actual GlitchTip.

Having both bandwidth and stanza IDs would be useful instead of only
ever capturing one. With both, message_id becomes redundant.

Use provided timestamp everywhere, don't generate a timestamp when one's
not provided and instead just rely on the event ID from redis, because
stream IDs are timestamps.

The more I tried to clean up and improve the use of ValueSemantics, the
more I came to the conclusion it would either have way too much
duplication or deduplicating would be too messy and complex.

- depend on Amolith's fork of Rantly. Fix has moved into upstream but
  requires jump to Ruby 3.3; should do eventually, but 3.3 isn't even
  packaged in our Guix commit.
- pin to same Rubocop version as sgx-jmp.
- pry-byebug

Some dependencies or other depend on them, but don't add them to their
Gemfile. However, they were removed from the stdlib in later Ruby
versions, so things break if we don't explicitly include them.

I refactored away how I was using it before committing
51d5802e2782f49f6ade16e36d0caf599a9020ef, then forgot to remove it.

Version 2.7.0 added TnOptions endpoints needed for port-out PIN
functionality. The lock file had 1.0.2 which predates that feature.

- in: inbound message-received
- out: outbound after successful Bandwidth API call
- thru: pass-through XMPP-to-XMPP messages
- delivered: message-delivered callback
- failed: message-failed callback

Implements: https://todo.sr.ht/~singpolyma/soprani.ca/517
Implements: https://todo.sr.ht/~singpolyma/soprani.ca/518

* '380-conditionally-show-command' of https://git.secluded.site/sgx-bwmsgsv2:
  refactor: only show port-out cmd if tn eligible

References: https://todo.sr.ht/~singpolyma/soprani.ca/380
Co-authored-by: Crush <crush@charm.land>

* '380-set-port-out-pin' of https://git.secluded.site/sgx-bwmsgsv2:
  chore: swap StandardError/RuntimeError, simplify
  feat: let users set their port-out PIN
  chore: update rubocop names
  test: rescue NameError
  chore: add missing webrick dependency