View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0012269 | Taler | exchange | public | 2026-11-21 10:28 | 2026-11-21 22:42 |
| Reporter | bot-oec [agent] | ||||
| Implementer | oec | ||||
| Priority | high | Severity | major | ||
| Status | implementing | ||||
| Target Version | 1.6 | ||||
| Summary | 0012269: The coin history has no entry with the blinded signatures of a reveal | ||||
| Description | For the owner of an old coin to recover the fresh coins of its melts, the first correct reveal has to append an entry with the blinded signatures of the revealed batch `γ` (for Clause-Schnorr also the R-values) to the old coin's history. The exchange follows protocol v46 as documented, which provides these signatures nowhere. Commit `5d5e2ef86` ("coin history: no ev_sigs in MELT entries", 0011916) removed the field `ev_sigs` from the history handler and from `get_coin_transactions.c`, where it was returned only for revealed melts (`CASE WHEN revealed THEN denom_sigs`). A revert of `5d5e2ef86` is not a fix. Its server-side part (history handler, `exchangedb/get_coin_transactions.c`, `free_coin_transaction_list.{c,h}`, `test_coin_history.c`) applies without conflicts and puts `ev_sigs` back into the `MELT` entry of revealed melts, while `test_exchange_api_{cs,rsa}` and `test_coin_history` still pass. But the reveal adds no `coin_history` row: the melt inserts that row (`refresh_insert_trigger`, `0002-refresh.sql`), while the reveal only sets `refresh.revealed`. The ETag is the largest `coin_history_serial_id`, and a request with `start` returns only entries after that offset. So a client that read the history between melt and reveal gets `304 Not Modified`, or no entries after its offset, and never learns the signatures. **Steps to reproduce:** 1. Melt a coin. 2. Read the old coin's history. 3. Reveal. 4. Read the entries after the offset of step 2: expected a `REVEAL` entry with the melt's `rc` and one blinded signature per fresh coin, gets no entries (also with `5d5e2ef86` reverted). The test `src/testing/test_exchange_api_history_reveal.c` does this; it fails on master at step 4 (expects a `REVEAL` entry, gets no entries after the offset). The test will be on branch `dev/oec/formalization-findings` on git.taler.net; a note here will name the commit. **Proposed fix:** in the reveal transaction, when `refresh.revealed` turns true, insert a `coin_history` row for the reveal (a migration); serve it as a `REVEAL` entry with `rc` and the signatures (and, for CS, the stored `cs_r_values`); parse it in the client library. The ETag and `start` then advance with the reveal. For Clause-Schnorr denominations, the signatures are enough only if the exchange also checks that each envelope's nonce is the one derived from the blinding seed (it does not today). The owner must also be able to obtain the R-values: from `/blinding-prepare` while the denomination is valid, or from `cs_r_values` in the history entry. **Found by:** the formalization of the Taler protocols; the rule violated is (E2) in `spec/refresh.md` and (E1) in `spec/history.md`: the first correct reveal appends an entry with the blinded signatures of the revealed batch (for CS also the R-values) to the old coin's history. | ||||
| Tags | No tags attached. | ||||
| related to | 0012264 | reviewing | Clause-Schnorr planchets are signed under the nonce the client chose; a repeated nonce exposes the denomination private key |
| Test: branch `dev/oec/formalization-findings` of exchange.git (git.taler.net), commit ed8f69fdf82b "testing: reveal entry in the coin history (0012269)", on master 7b658d8da. test_exchange_api_history_reveal_{cs,rsa} melts a coin, reads its history, reveals, and reads the entries after the first read: expected a REVEAL entry with the melt's rc and one blinded signature per fresh coin, gets no entries. The client half of the proposed entry comes with the test: the type TALER_EXCHANGE_CTT_REVEAL and its parser in libtalerexchange (a REVEAL entry carries rc and ev_sigs, no amount), and the testing commands TALER_TESTING_cmd_coin_history_mark and TALER_TESTING_cmd_coin_history_reveal_entry. Fails on master at history-reveal-entry. | |
| Fix proposed for review: exchange branch dev/oec/fix-12236, commits d5d94bac8 (the test of the note above, with the client parser for the new entry) and ca11dae16 (the REVEAL entry: trigger on refresh_reveal in migration exchange-0021, transaction list, GET /coins/$COIN_PUB/history, auditor checks, database tests). test_exchange_api_history_reveal_{cs,rsa} passes; the protocol side is #12236. | |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2026-11-21 10:28 | bot-oec | New Issue | |
| 2026-11-21 10:28 | bot-oec | Relationship added | related to 0012264 |
| 2026-11-21 10:28 | bot-oec | Implementer | => oec |
| 2026-11-21 10:28 | bot-oec | Reviewer | => Christian Grothoff |
| 2026-11-21 10:28 | bot-oec | Release signed by | => Christian Grothoff |
| 2026-11-21 10:28 | bot-oec | Deployed by | => Christian Grothoff |
| 2026-11-21 10:28 | bot-oec | Verifier | => oec |
| 2026-11-21 10:28 | bot-oec | Communicator | => CM_7 |
| 2026-11-21 10:47 | bot-oec | Note Added: 0030669 | |
| 2026-11-21 12:27 | bot-oec | Note Added: 0030693 | |
| 2026-11-21 16:14 | oec | Implementer | oec => |
| 2026-11-21 22:31 | Christian Grothoff | Locked by | => Christian Grothoff |
| 2026-11-21 22:31 | Christian Grothoff | Status | new => planning |
| 2026-11-21 22:31 | Christian Grothoff | Implementer | => oec |
| 2026-11-21 22:31 | Christian Grothoff | Priority | normal => high |
| 2026-11-21 22:31 | Christian Grothoff | Target Version | => 1.6 |
| 2026-11-21 22:31 | Christian Grothoff | Locked by | Christian Grothoff => |
| 2026-11-21 22:31 | Christian Grothoff | Status | planning => implementing |
| 2026-11-21 22:42 | Christian Grothoff | View Status | private => public |