Address subscription does not scan blocks or transactions.

[santoslluis]

Hello,

I tried to do bx watch-address with libbitcoin-explorer (bx) and with my own implementation in PHP against the ZMQ API, but I didn't receive any update from the server when there is a transaction.

perhaps is a bug, can anyone confirm this?

L.

[evoskuil]

There may be an issue with the server. Address subscription is not part of the network test matrix and as such a regression may evade detection. I plan to look into this shortly, as another person has reported a similar problem.

[evoskuil]

Are you running BX against your own server (BS)? Is the server configured to respond to queries (it isn't by default)?

[santoslluis]

Yes, I tried with my own server, with obelisk.airbitz.co and another from the dark wallet team.

the other commands are working properly, the only that doesn't work is address.subscribe

I have the parameter 'queries_enabled' set to true.

[evoskuil]

OK, forgot to mention that you would also need to have set:

[network]
# Request that peers relay transactions, defaults to false.
relay_transactions = true

[santoslluis]

I've set this parameter to true and is still not working the address.subscribe command

[evoskuil]

I've made a couple of fixes and improvements to the client-server API subscription implementation in the server. I'm in the process of merging from my repo in the various affected repo version2 branches. I'll be tagging a release in the next couple of days. Thanks for reporting this issue!

[evoskuil]

Fixed in #148 and dependent repo version2 branches.

[evoskuil]

Analysis:

Transactions that are discovered via block acceptance are not processed. This may have been a regression and was a simple fix. Block accept notifications are in addition to what may have previously been sent via mempool acceptance, so typically you should get 1 mempool notification and 1 block acceptance notification. Reorgs and mempool overflow with tx resend can of course cause variations.

Notification from mempool txs is inconsistent. I believe this was a regression stemming from an incorrect virtual method override. It seems that in some builds it works as expected and in others, not at all. I've changed this technique from virtual handler override to mempool acceptance subscription. This will mirror the block acceptance technique (reorg subscription).

[santoslluis]

Well, I've recompiled, tested and now address.subscribe works fine.

Thanks!

[evoskuil]

Great, thanks for the feedback and for reporting the issue!