Initialize persistent stores regardless of enable_transport

[wet-bulb]

Fixes announce handling when transport is disabled

Transport::start() only initialises the path table, known-destinations store, and packet hashlist inside if (Reticulum::transport_enabled()). So if you run an instance with transport disabled, those stores never get set up, and every inbound announce fails:

[ERR] remember: failed to store identity for <hash>
[ERR] Failed to add destination <hash> to path table!

No path ever gets added. The instance announces fine, logs nothing else, and just can't address a peer.

Easy to reproduce: with transport disabled, all four test_interop scenarios (packet, link, request, resource) fail the same way. (Worth noting the test_interop senders all set transport_enabled(false) themselves.) With the stores initialised, they pass bidirectionally against Python RNS 1.4.2 and 1.1.9. It's not version-specific either, 1.2.9 reproduces it too.

The fix: every instance needs an identity cache, a path table, and a packet hashlist to address anything, so those three initialisations move out of the gate. The tunnel table read and the probe destination stay inside it, since those really are transport-only.

Nothing changes for transport-enabled instances. The same three initialisations run in the same relative order, just before the if instead of inside it.

Tested on a native host build (pio -e native17, Ubuntu, Python RNS from pip). Haven't run it on hardware yet.

Opened at the request in #74. The diff looks bigger than it is, most of it is re-indentation from dropping a nesting level. With whitespace hidden (?w=1) it's about fourteen lines.

[attermann]

Thanks for the PR. Would you mind checking your changes against the Python reference implementation and noting where in that code this initialization takes place?

…

On Sun, Aug 2, 2026 at 1:15 PM gelly ***@***.***> wrote: *Fixes announce handling when transport is disabled* Transport::start() only initialises the path table, known-destinations store, and packet hashlist inside if (Reticulum::transport_enabled()). So if you run an instance with transport disabled, those stores never get set up, and every inbound announce fails: [ERR] remember: failed to store identity for <hash> [ERR] Failed to add destination <hash> to path table! No path ever gets added. The instance announces fine, logs nothing else, and just can't address a peer. Easy to reproduce: with transport disabled, all four test_interop scenarios (packet, link, request, resource) fail the same way. (Worth noting the test_interop senders all set transport_enabled(false) themselves.) With the stores initialised, they pass bidirectionally against Python RNS 1.4.2 and 1.1.9. It's not version-specific either, 1.2.9 reproduces it too. The fix: every instance needs an identity cache, a path table, and a packet hashlist to address anything, so those three initialisations move out of the gate. The tunnel table read and the probe destination stay inside it, since those really are transport-only. Nothing changes for transport-enabled instances. The same three initialisations run in the same relative order, just before the if instead of inside it. Tested on a native host build (pio -e native17, Ubuntu, Python RNS from pip). Haven't run it on hardware yet. Opened at the request in #74 <#74>. The diff looks bigger than it is, most of it is re-indentation from dropping a nesting level. With whitespace hidden (?w=1) it's about fourteen lines. ------------------------------ You can view, comment on, or merge this pull request online at: #75 Commit Summary - a84fb1d <a84fb1d> Initialize persistent stores regardless of enable_transport File Changes (1 file <https://github.com/attermann/microReticulum/pull/75/files>) - *M* src/microReticulum/Transport.cpp <https://github.com/attermann/microReticulum/pull/75/files#diff-ca58be9981a1b8bcfeb339e74a13ed1e2ab51bd340ba12002732d61b0a8215f4> (82) Patch Links: - https://github.com/attermann/microReticulum/pull/75.patch - https://github.com/attermann/microReticulum/pull/75.diff — Reply to this email directly, view it on GitHub <#75?email_source=notifications&email_token=ACXDLGNTHGAQE7XG3GXS7KL5H6HM5A5CNFSNUABEM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UF42DCOBZGU2TANZSG6THEZLBONXW5KTTOVRHGY3SNFRGKZFFMV3GK3TUVRTG633UMVZF6Y3MNFRWW>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACXDLGOOAIK5U6DF5HKTXMD5H6HM5AVCNFSNUABFKJSXA33TNF2G64TZHM3TAMJVGYYDGNJTHNEXG43VMU5TKMBUGQYDMMBVHAZKC5QC> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[wet-bulb]

three places, all on current master:

known destinations: Reticulum.py:344 calls RNS.Identity.load_known_destinations() in Reticulum.init, two lines before RNS.Transport.start(self). not gated on anything.

path table: Transport.py:115, path_table = {} as a class attribute, so it always exists. only the disk restore is gated, at :306, under the same # Load transport-related data comment this code has.

packet hashlist: same shape. :107 packet_hashlist = set() unconditional, disk restore gated at :243.

so Python's split is that the structures always exist and only restoring their contents from storage is transport gated.

that split doesn't map cleanly onto microStore. FileStore::init() attaches the store and makes persisted records available in one step, and its third parameter is clearOnInit rather than anything that would let you attach without reading. the store is the structure, so there's no equivalent of "empty dict now, fill it later".

so this PR matches Python exactly for the first: known destinations. and for the other two it always restores where Python restores only when transport is on. that difference seemed acceptable to me on a constrained radio node (transport disabled scenario), where reloading known paths after a reboot saves announce and path request traffic, but it is a difference and you may not want it.

if you'd rather the port mirror the reference precisely, that needs a microStore change to separate attach from load. happy to do it that way instead.

[drupol]

It would be nice to link to the source on GitHub.

[wet-bulb]

good point

permalinks, pinned to b48b96e:

known destinations: Reticulum.py:344
path table: Transport.py:115, disk restore at :306
packet hashlist: :107, disk restore at :243