Skip to content

Promote dev to main: exporter express fix, web3 2.0.0, CI - #64

Merged
moscowchill merged 3 commits into
mainfrom
dev
Nov 9, 2026
Merged

moscowchill merged 3 commits into
mainfrom
dev

Conversation

@moscowchill

Copy link
Copy Markdown
Contributor

Promotion with a merge commit. Brings #61 (express 4.22.3 in the contract exporter: body-parser and proxy-addr advisories), #62 (@theqrl/web3 2.0.0) and #63 (CI runs each change once). Clears the zod, body-parser and proxy-addr advisories that the daily security sweep reports on main since 24 Oct. No deploy is part of this PR.

body-parser 1.20.9 (GHSA-27r3-jxmh-p828, a compressed body with trailing
bytes stops the process) and proxy-addr 2.0.8 (GHSA-jqcg-44mw-7w3h, trust
check bypass with IPv4-mapped IPv6). The exporter is not deployed today;
this keeps it safe to run.
* chore(deps): upgrade @theqrl/web3 to 2.0.0 in the tooling and exporter

The root tooling and monitoring/contract-exporter move from ^1.0.3 to
^2.0.0. A caret range never crosses a major, so both needed the edit.
Every @theqrl/web3* package resolves to 2.0.0 with one copy each, and
web3-qrl-iban and web3-providers-ipc leave both trees with the umbrella.

Root overrides:
- Drop the $-overrides for @theqrl/mldsa87, @theqrl/wallet.js and
  @theqrl/qrl-cryptography. 2.0.0 pins mldsa87 2.2.0, wallet.js 6.3.0
  and qrl-cryptography 0.3.3 itself, and npm ls shows a single copy of
  each.
- Keep ws ^8.21.0 as the ws security floor; it lifts the providers-ws
  pin from 8.22.0 to 8.22.1.
- Keep tmp and follow-redirects. Nothing in the current tree resolves
  them; they stay as floors for a future dependency.

The exporter had no crypto overrides and now gets mldsa87 2.2.0,
wallet.js 6.3.0 and qrl-cryptography 0.3.3 (from 2.1.3, 6.2.3, 0.3.0).
Its qs override is unrelated and stays.

npm audit: root 17 high to 0. Exporter 22 (17 high, 5 moderate) to 5
moderate, the qs, express, body-parser, minimatch and nodemon entries it
already had. The 17 highs were the web3 chain over the validator's
nested zod 3.22.3; 2.0.0's validator ships zod 4.6.7.

No source change is needed here: the scripts sign type 2 with fee-market
fields, deployer seeds are already normalized to 0x, and the finality
proposal ids encode identically under both versions.

* test(tooling): drive the exporter and deployer loader through web3 2.0.0

The existing suites hand both modules a fake web3, so nothing showed
that they work with the installed client.

- native-monitoring: the exporter scrapes one block through the real
  @theqrl/web3 client and a local JSON-RPC endpoint that echoes request
  ids. 2.0.0 refuses a reply with another id, decodes addresses as
  checksummed Q addresses and decodes every word strictly; the finality
  and portfolio bindings go through those decoded addresses. The case
  passes on 1.0.3 as well, so the exporter's behaviour is unchanged.
- loadDeployer: a bare, 0X and padded 0x seed each reach web3 as the 0x
  form and load the same account. 2.0.0 refuses the bare hex with
  InvalidSeedError, which the case also pins; 1.0.3 accepted it.

* fix(frontend): upgrade to @theqrl/web3 2.0.0 and filter logs by its 64-byte topic

The staking frontend moves from ^1.0.3 to ^2.0.0. Every @theqrl/web3*
package resolves to 2.0.0 with one copy each. The exact overrides for
@theqrl/mldsa87 (2.2.0), @theqrl/wallet.js (6.3.0) and
@theqrl/qrl-cryptography (0.3.3) are dropped: 2.0.0 pins the same
versions, and @qrlwallet/connect's ^2.1.1 mldsa87 range dedupes to the
same copy. The overrides for brace-expansion, esbuild, js-yaml, nanoid,
postcss and socket.io-parser pin packages outside web3 and stay.

npm audit: production 17 high to 0; the full tree 32 to 15, the eslint,
minimatch, brace-expansion, source-map-js and vite entries it already
had.

Activity feed: abi.encodeEventSignature returned the bare 32-byte
keccak hash in 1.x and returns the 64-byte VM64 topic in 2.0.0, the
hash in the high 32 bytes and zeros in the low 32 bytes. nativeEventTopics
required the 32-byte form and padded it, so on 2.0.0 every activity
query threw "Invalid native event filter": a connected account saw
"Activity history is unavailable from this RPC" and an empty refund
list. It now takes the 64-byte topic and requires the zero low half.
The qrl_getLogs request is byte-for-byte the one 1.0.3 sent. A 32-byte
signature fails closed before any request, so a future change in the
library's topic layout shows up as an error.

The comment above the raw request named a 1.x validator limit that
2.0.0 lifts; it now records that history and why the request stays raw.

Tests: the topic fixtures use the 64-byte form, the decoder test checks
that the high half is keccak256 of the event signature, and new cases
refuse the 32-byte width, a low-aligned hash and a non-zero low half.

* test(frontend): pin the checksummed beneficiary web3 2.0.0 decodes

Since @theqrl/web3 2.0.0, decoded addresses come back with QIP-55
checksum casing; 1.x returned the lowercase body. The refund list checks
each getPending beneficiary with isQrlAddress, which accepts a
mixed-case body only when it matches the app's own SHAKE-256 checksum.
The existing case used a digits-only address, where casing never shows.

The new case decodes a getPending record with the installed web3, for an
account with hex letters, checks that the beneficiary is mixed case, and
expects the refund list to hold the deposit with no activity error.
Every commit on a branch with an open pull request ran twice, once for the push
and once for the pull request: 26 of the last 40 runs were pushes.

- Pull requests into dev and main run the suite, plus a weekly run on dev and
  manual dispatch.
- Job timeout, superseded pull request runs cancelled, actions pinned by commit
  at current majors.
@moscowchill
moscowchill merged commit f23e19a into main Nov 9, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant