Public: anyone can read this; the crew can change it.
/talk?about= names its subject: one line above the composer, only for what the viewer can read #10497
/talk?about=owner/repo and /talk?about=owner/repo#N now show one line above the composer: "About owner/repo" or "About owner/repo#N: ", with the title escaped. The 38 tests in test/talk-page.test.ts pass and bun x tsc --noEmit is clean.
stopped parked by the crew deploy df958c7
3 earlier comments
/talk?about=owner/repo and /talk?about=owner/repo#N now show one line above the composer: "About owner/repo" or "About owner/repo#N: <title>", with the title escaped. The 38 tests in test/talk-page.test.ts pass and bun x tsc --noEmit is clean. bun scripts/design-lint.ts printed "no colour, size or component added outside the system" and I didn't check its exit code.
- Parsing:
parseAbout(text)insrc/web/pages/talk.tsis a pure function, string in,{owner, repo, number?}or null out. It rejects anything over 140 characters,../x,a/b#x,a/b#0and a trailing space. - Lookup:
aboutLineruns only on a valid parse. It requirescanReadon the repository, and for an issue or pull alsomayViewId, so sealed reports stay hidden. A missing, private or unreadable subject draws nothing, so the page is byte-identical to a repository that doesn't exist. #Nin the URL: a literal#is a browser fragment and never reaches the server, so the number has to arrive encoded as%23N. The tests useencodeURIComponent. The later link-changing parts of #10326 will need to write it that way.- Tests:
- A readable issue shows its escaped title, and a repository shows its name.
- A private repository and its issue name nothing for another signed-in person, and the response equals the one for a repository that doesn't exist. The owner still sees the title.
- The malformed values (../x, a/b#x, 300 characters) leave a page identical to plain /talk.
- Signed out:
/talkredirects to sign-in and echoes the request URL innext. Nothing is looked up, so private and missing repositories behave the same. My test asserts that redirect, not a byte-equal page, since there is no page to compare. - Not done: no links point at the new parameter yet and the line has no styling of its own, as the issue asked.
- Push hook: the pre-push hook demanded a commit message containing a "Try:" line. I added an empty commit for it: open
/talk?about=mason/forgesigned in and read the line above the composer.
This pull's /talk?about= change is untouched. The two failing tests were stale on main, not broken by this pull. I ran them, test/talk-page.test.ts and bun x tsc --noEmit locally and they pass. I did not re-run the full suite; the forge's check does that on this push.
What changed
scripts/unreached-allow.json: I removedsrc/kernel/why_now.ts. That file no longer exists, so the entry made "an allowed module that has become reached leaves the list" fail.test/child-quota.test.ts: a small forge now holds 2048 MiB, up from 512 MiB, and an earlier test in the same file already holds one. I raised the memory limit in "room for small but not large is refused only for large" from 2000 to 5000 MiB. A second small forge now fits, and a large one (8192 MiB) is still refused. The old limit was wrong for the new size, so this is a test fix and not a code fix.
What the pull still does: /talk?about=<repo> shows its subject in one line above the composer, and only for repositories the viewer can read.
How to try it: open /talk?about=mason/forge in a signed-in browser and read the line above the composer.
🤖 Generated with Claude Code
I changed no code for this failure. I merged current main into the pull and pushed it. Locally, the failing tests I could run pass, and the cell and sandbox failures are the ones main just fixed.
The eight failures don't touch /talk?about=. They fall into three groups:
- Child cell and crew-shell sandbox failures: the child-forge cell tests, the "shell sandbox denies the whole data directory" tests and "the member actor … crew shell" failed. These are the areas two new
maincommits changed:636aa0355(a child's crew-run temp directory) and0dfe641b7(a child's cell hiding the parent's home). The merge brings both in. I can't run cell or unix-socket suites in this sandbox, so the forge's check on this push is the real test of these. - Environment: the check log shows the machine's
/tmpat 243 of 256 MiB free. Both tests in the next group may have failed from that. - Two UI tests: "a repository name with markup in it is escaped in the select, never raw" and "a pin's cart > a captured cart plays sandboxed" both pass here. They are in
test/feed-composer-language.test.tsandtest/build-cart.test.ts.
After the merge, test/talk-page.test.ts, test/unreached.test.ts and test/child-quota.test.ts pass and tsc is clean.
The pull itself is unchanged. /talk?about=<repo> puts its subject in one line above the composer, only for repositories the viewer can read. To try it, open /talk?about=mason/forge signed in. The earlier commit abce63fc4 also removed a stale allow-list entry and corrected the memory limit in the child-quota test.
If the cell and sandbox tests are still red after this push, the cause is outside this pull.
🤖 Generated with Claude Code
Parked by the crew: the check stayed red through 2 automatic tries
/talk?about=owner/repoand/talk?about=owner/repo#Nnow show one line above the composer: "About owner/repo" or "About owner/repo#N: <title>", with the title escaped. The 38 tests intest/talk-page.test.tspass andbun x tsc --noEmitis clean.bun scripts/design-lint.tsprinted "no colour, size or component added outside the system" and I didn't check its exit code.parseAbout(text)insrc/web/pages/talk.tsis a pure function, string in,{owner, repo, number?}or null out. It rejects anything over 140 characters,../x,a/b#x,a/b#0and a trailing space.aboutLineruns only on a valid parse. It requirescanReadon the repository, and for an issue or pull alsomayViewId, so sealed reports stay hidden. A missing, private or unreadable subject draws nothing, so the page is byte-identical to a repository that doesn't exist.#Nin the URL: a literal#is a browser fragment and never reaches the server, so the number has to arrive encoded as%23N. The tests useencodeURIComponent. The later link-changing parts of #10326 will need to write it that way.- A readable issue shows its escaped title, and a repository shows its name. - A private repository and its issue name nothing for another signed-in person, and the response equals the one for a repository that doesn't exist. The owner still sees the title. - The malformed values (
../x,a/b#x, 300 characters) leave a page identical to plain/talk./talkredirects to sign-in and echoes the request URL innext. Nothing is looked up, so private and missing repositories behave the same. My test asserts that redirect, not a byte-equal page, since there is no page to compare./talk?about=mason/forgesigned in and read the line above the composer.Closes #10494