Discourse

Recent Commits

• Kris (04 Jul 26)

  UX: minor padding fix for mobile PM topic map (#41445) Too much left padding on mobile in the PM topic map before <img width="350" alt="image" src="https://github.com/user-attachments/assets/fb5354a8-a01c-4d7d-9e18-a122ea51594c" /> after <img width="350" alt="image" src="https://github.com/user-attachments/assets/f39cb8fa-6035-43e2-ba3a-e97b8eb75a83" />

• Kris (03 Jul 26)

  FIX: avoid double arrow key nav for select-kit-filter (#41443) Currently when you search for a tag in the composer, it requires two arrow key presses to reach the second option, even though the first option is already highlighted. This is because the create option (first row) is pre-highlighted... so arrow down in select-kit calls `this.selectKit.highlightFirst()` and you end up re-highlighting it on the first arrow press. This checks if a row is pre-highlighted to avoid `highlightFirst` and instead fires `highlightNext` (and the opposite for up arrow cases)

• Jordan Vidrine (03 Jul 26)

  UX: Remove new height variables from foundation modernization (#41440)

• Michael Brown (03 Jul 26)

  DEV: remove x-real-ip references (#41371) Discourse (rails) doesn't actually use the `x-real-ip` header, so remove it from the sample nginx configuration. Discourse was also logging it in `discourse/lib/discourse_logstash_logger.rb`, but it doesn't actually mean anything to the application so there's no point in doing so.

• Gabriel Grubba (03 Jul 26)

  SECURITY: Filter private subcategory counts from category listings (#41384) The category list aggregate topic counts included direct subcategories without checking whether the current guardian could see them. Before we the only possible information that could be gotten was: - private subcategory topic counts - recency buckets for topic creation: day/week/month/year/all-time Co-authored-by: discourse-patch-triage <272280883+discourse-patch-triage[bot]@users.noreply.github.com>

• Renato Atilio (03 Jul 26)

  FIX: Keep quotes around wrap bbcode attribute values with spaces (#41429) The rich text editor's wrap serialize/parse helpers are meant to mirror the server bbcode parser, but they never handled quoting: `serializeAttributes` emitted bare `key=value` pairs and `parseAttributesString` split on whitespace. So a value containing spaces such as `repo-name="Homepage Feature"` was rewritten as `repo-name=Homepage Feature` on the next markdown/RTE toggle, breaking the bbcode. `serializeAttributes` now wraps values that contain whitespace in double quotes, matching how `[quote=...]` already serializes. `parseAttributesString` reconstructs the tag and delegates to the markdown-it `parseBBCodeTag`, so it reads quoted values back in full parity with the server (every supported quotation mark, not just double quotes), keeping the round-trip lossless.

• Penar Musaraj (03 Jul 26)

  DEV: Apply different first-day limits for TL0 and TL1 (#41433) Two changes here: - the first-day limits clock now starts when user account is created (previously, it was starting when the first post was created) - TL1 gets its own limits, by default 6 topics and 30 replies, configurable via its own site settings `tl1_max_topics_in_first_day`, `tl1_max_replies_in_first_day` I'm not thrilled to add yet another site setting here, but this change is useful, it allows better separation between TL0 and TL1 in a critical first day of use. Context in https://meta.discourse.org/t/bootstrap-mode-is-legacy/402468/17?u=pmusaraj

• Penar Musaraj (03 Jul 26)

  DEV: auto-silence only if a post is queued (#41435) A new user could previously get in a permanent silenced state by simply uploading 2 or more media items in their first post by uploading quickly and not adding any content. Our code would silence them (via fast typing detection) and not queue the reviewable because the post would be rejected, from `newuser_max_embedded_media`. This fix ensures that the silence only happens if the post is queued for review.

• Rafael dos Santos Silva (03 Jul 26)

  DEV: Add `full-page-refresh-on-navigation` value transformer (#41436) Previously, when a new client version was deployed, the next navigation always became a full page load, which destroyed long-lived client state — for example dropping users out of an ongoing voice call. This change wraps that decision (in `DiscourseURL#routeTo` and `DNavItem`) in a new `full-page-refresh-on-navigation` value transformer, so plugins can defer the reload while such state is active; the refresh then happens on the first navigation after the transformer stops deferring it. Co-authored-by: David Taylor <[email protected]>

• David Taylor (03 Jul 26)

  DEV: Drop (almost)-unused `TopicEntrance` component (#41424) We stopped using this UI in most topic lists as part of the raw-hbs -> glimmer conversion, since it was seldom-used, and simple links are a better UX. However, the implementation remained, and it was still accessible in some very specific places (e.g. user-activity/topics on mobile devices). This commit strips out the implementation, so we have consistent behavior across all topic lists.

• David Taylor (03 Jul 26)

  DEV: Drop unused `decorateCooked` from discourse-details (#41426) This has been a no-op since 8f2f9e6afa0 in 2020, when we removed the JS polyfill of `<details>`

• David Taylor (03 Jul 26)

  DEV: Convert caret-position from JQuery to vanilla JS (#41428) This was verified by a/b testing a large number of scenarios against old and new implementations, and confirming that they returned precisely identical coordinates.

• David Taylor (03 Jul 26)

  DEV: Convert reactions animations from JQuery to native (#41427)

• David Taylor (03 Jul 26)

  DEV: Drop jquery in patreon plugin (#41425) This was doing direct manipulation of ember-rendered DOM, which is risky. This commit changes it to be a simple non-animated closure, which is safer, and more in line with other Discourse UX.

• David Taylor (03 Jul 26)

  DEV: Remove jquery usage from `lib/url` (#41422)

• David Taylor (03 Jul 26)

  DEV: Remove jquery from select-kit-helper and category-badge-test (#41421) - The 'selectAll' path in the select-kit-helper is completely unused, so we can just drop it - Update HTML parsing in category-badge-test

• discoursebot (03 Jul 26)

  FEATURE: Bump ai_bot_enable_docked_composer upcoming change to beta (#40922) <!-- upcoming-change-status-pr:ai_bot_enable_docked_composer --> This automated PR moves `ai_bot_enable_docked_composer` from `alpha` to `beta` after 14+ days without a status change. - Last status change commit: [`b3b561e5fad412c038e222499ebe22050b4a8de4`](https://github.com/discourse/discourse/commit/b3b561e5fad412c038e222499ebe22050b4a8de4) - Last status change date: `2026-05-04T09:26:32-07:00` - Settings file: `plugins/discourse-ai/config/settings.yml` - Original author: Keegan George (<[email protected]>) - Original PR: #39708

• Régis Hanol (03 Jul 26)

  UX: Normalize automatic membership email domains (#41434) Previously, entering an automatic membership email domain that included an `@` (such as `@acme.io` or a pasted email or URL) warned the admin and blocked the save until they hand-edited it down to a bare domain. This change quietly normalizes those entries to the bare domain — stripping any scheme, path, port, `@` prefix, casing, and surrounding whitespace — so common paste mistakes just work, while the server still rejects genuinely invalid domains on save. Follow up to #41233

• Gerhard Schlager (03 Jul 26)

  MT: Let a generated model declare its conflict strategy in the schema DSL The per-table merge policy reads `conflict_strategy` off each model, but only the hand-written `upload` model could declare it — a generated model always got a plain `INSERT` because the generator has no notion of a conflict clause. This adds a `conflict_strategy :ignore` directive to the table DSL and threads it through to `ModelWriter`, so a generated model can opt into `INSERT OR IGNORE`: the writer emits both the `OR IGNORE` SQL and a `self.conflict_strategy` method, which is what `IntermediateDB.conflict_strategy_for` reads at run time. It defaults to `:raise` and only emits anything for `:ignore`, so every existing generated model regenerates byte-for-byte. No table uses it yet — `uploads` stays a manual model — but the option now lives in the schema config where the rest of a table's shape is declared.

• Gerhard Schlager (03 Jul 26)

  MT: Only use INSERT OR IGNORE for tables that want it in the shard merge The sharded write path (#41262) merges every shard table with `INSERT OR IGNORE`. That's right for `uploads`, but for every other table it silently swallows a genuine cross-step/cross-shard duplicate row that the single-writer path would have raised on — so a concurrency bug that doubles a key becomes silent data loss with no error and no baseline to diff against. This scopes the `OR IGNORE` back to just the tables that actually want it, derived from the models rather than a hardcoded table name: - `IntermediateDB.conflict_strategy_for(table)` resolves a table's model and reads its declared `conflict_strategy`, defaulting to `:raise`. - Only `upload` declares `:ignore` (its rows are content-identical for a given id-hash, so first-writer-wins is intended). A future `OR IGNORE` model flips its own merge clause with no change to the merge code. - `Connection#merge_database` takes `dedupe_tables:` and picks the clause per table; a plain-table conflict raises, wrapped with the table name, and the error surfaces through `Consolidator#drain` as a run failure. This restores the earlier behaviour where a genuine duplicate on a plain table is an error, not a dropped row.

• Gabriel Grubba (03 Jul 26)

  FIX: Ignore title parameter during public signup (#41381) Public account creation accepted `title` from non-admin signup requests, allowing a new user to assign themselves an arbitrary visible profile title before normal title authorization checks could apply. Ignore `title` for non-admin account creation, matching the existing treatment of primary/flair group parameters. Admin API user creation remains unchanged. Co-authored-by: discourse-patch-triage <272280883+discourse-patch-triage[bot]@users.noreply.github.com>

• Gabriel Grubba (03 Jul 26)

  DEV: outlet for thread actions (#41382) This adds an outlet for both thread actions in drawer and full page mode

• Joffrey JAFFEUX (03 Jul 26)

  FEATURE: adds user-search trigger to workflows (#41423) Adds a new trigger node: user-search. This will trigger when a user is doing a non empty search through the UI. To make it possible: - a new DiscourseEvent trigger has been created: user_search - the workflows cache has been improved/simplified to allow for different kind of triggers/nodes

• David Taylor (03 Jul 26)

  FIX: Prevent error when a reviewable claim is broadcast for a topicless reviewable (#41418) Previously, claiming a topic in the review queue threw `Cannot read properties of null (reading 'id')` in every `ReviewableItem` whose reviewable had no associated topic (e.g. `ReviewableUser` or a queued new topic), because `_updateClaimedBy` dereferenced `this.reviewable.topic.id` on the `/reviewable_claimed` broadcast. This change compares against the existing `topicId` getter, which safely resolves the topic id and lets topicless reviewables ignore the broadcast instead of erroring.

• David Taylor (03 Jul 26)

  DEV: Use `sinon` to stub `capabilities.touch` (#41419) Using `Object.defineProperty` means that it persists and affects other tests in the suite Followup to 64d03062de

• chapoi (03 Jul 26)

  UX: fix topic title edit aligments (#41417) The translation info increased the size of the `edit-title__wrapper`, which in turn broke the absolute positioning (necessary evil) of the AI helper button. This commit moves `editing-localization-indicator` a parent up, to `edit-topic-title`, so it's just part of the flex layout and no longer interferes with the wrapper. | Then | Now | |--------|--------| | <img width="822" height="203" alt="CleanShot 2026-07-03 at 11 44 09" src="https://github.com/user-attachments/assets/ce525ab1-2028-43a3-86cd-61be5ae66779" /> | <img width="822" height="203" alt="CleanShot 2026-07-03 at 11 43 35" src="https://github.com/user-attachments/assets/58aa1c28-2f65-469c-9769-08279062ad16" /> |

• Natalie Tay (03 Jul 26)

  SECURITY: Restrict user action removal messages (#41405) Scope UserAction removal MessageBus publishes to the affected user so unrelated clients cannot receive removal metadata.

• Gerhard Schlager (29 Jun 26)

  MT: Run conversion steps concurrently and split the heavy ones across cores This is the big one I've been building up to across the step-concurrency series: the converter now uses every core instead of grinding through one step at a time on a single CPU. The win scales with the number of cores, so the figure depends on the machine. On my laptop, running all the currently implemented Discourse converter steps dropped from 102s to 18s, about 5.5x faster, with all cores busy the whole time. Two things were slow before: steps ran one after another, and the few genuinely large steps (topic_users above all) ran single-threaded even when nothing else was happening. This PR tackles both. Concurrent steps. A dependency-aware scheduler runs independent steps at the same time, each with its own source connection. It's event-driven: as soon as a step finishes, it frees its fork and the next ready step starts, so the cores don't sit idle between steps. How many run at once is bounded by the cores it can use, and `--max-parallel-steps` lowers it. A pull model for workers. Instead of the parent reading every row and streaming items down a pipe to the workers, each worker now opens its own source connection and reads its own slice directly. Only progress goes back over the pipe, and only once every thousand items, so it's never the bottleneck. The parent's single-threaded read was the real ceiling, and this removes it, along with a fair bit of machinery (the old worker pool, the item serialization, the handshake). Partitioning the heavy steps. A step opts in with `partition_by`, and the scheduler splits it across forks, each reading one chunk of the key range. I kept it general: a dense numeric key is divided into even chunks straight from a cheap MIN/MAX, while a sparse numeric key or a text/UUID or composite key falls back to a sorted-key scan so every chunk holds a similar number of rows whatever the key type. The dialect-specific SQL lives in the Postgres adapter, so other sources can fill in their own later. One caveat I documented on `partition_by`: only use it on steps whose processing is order-independent, since the forks run concurrently and their output is merged. Sharded writes. Each worker writes to its own SQLite shard, and a background consolidator folds finished shards back into the run database off the step's critical path, so a step no longer lingers at 100% while its merge runs. Reliability. A bad row or a worker that dies mid-step no longer hangs the run or takes down the other steps running at the same time: failures are caught and surfaced per step. Debugging. Forks make a debugger awkward, so there's a `--no-fork` flag: it runs each step inline in the main process, one at a time, so a breakpoint in a step's `process` stops where you can actually use it. The data path is the same (it still writes a shard the consolidator merges), only the fork is gone. Output order is no longer deterministic with concurrency, so to check correctness I added a small dev script under `migrations/tooling/scripts` that compares two IntermediateDBs order-insensitively and used it to confirm a parallel run produces the same data as a serial one. The gem suites are green.

• David Taylor (03 Jul 26)

  FIX: Stop asset-processor build from loading tsconfig (#41415) Our tsconfig is designed for development use only, and is not needed for a successful asset-processor build. In most cases, it didn't cause a problem. However, if someone does `rm -rf` on a core plugin, then it causes core's tscconfig to become invalid, which then breaks the asset-processor build.

• Discourse Translator Bot (03 Jul 26)

  I18N: Update translations (#41407)