Skip to main content
← Back to list
01Issue
BugClosedSwamp ClubPublic
Assigneeskeeb

Relationships

#1473 Profile edit broken on Safari

Opened by michael · 7/29/2026

Steps to reproduce:

  1. Use Safari
  2. Log into the website
  3. Go to Account -> Profile
  4. Click edit

Expected results:

You can update your profile

Actual results:

Nothing happens, looking at the dev console nothing happens

Additional info: it works on Chrome. Also it looks like the Lab -> Create issue is broken similarly.

02Bog Flow
OPENTRIAGEDIN PROGRESSCLOSED+ 1 MOREASSIGNED+ 2 MOREREVIEW+ 2 MORECODE_CONFORMANCE_REVIEW

Closed

8/14/2026, 10:30:16 PM

No activity in this phase yet.

03Sludge Pulse
keeb assigned keeb8/14/2026, 10:03:44 PM
Editable. Press Enter to edit.

keeb commented 8/14/2026, 10:30:16 PM

@michael thanks for this one — and sorry it sat. I couldn't reproduce it, and I went at it hard enough that I want to show the work rather than just say "works for me", because if you can still hit it the evidence below tells us exactly where to look next.

What I drove. A real WebKit 26.4 engine (the Safari 26.4 engine, headless, via Playwright) against both the current branch on a local server and against production, with headless Chromium as a control in every single run.

  1. Both surfaces work. Signed in, on /u/<name>: the Edit button is present, clicking it opens #profile-editor and it is visible. On /lab: "+ New" is present and clicking it opens the create form. Identical in Chromium. No page errors, no console errors, in either engine on either surface.

  2. Not a hydration failure. On production, the extension-registry island hydrated under WebKit — typing mongodb narrowed 24 results to 3 and pushed q= into the URL, which only client-side code does — with zero console errors or warnings on the load. So the shipped client bundle parses and runs in Safari.

  3. Not a scrollIntoView divergence, which was my best remaining theory and needed its own test. The Edit button dispatches an event and then, 50ms later, smooth-scrolls to the editor — which sits at the bottom of a long dossier. If that scroll silently failed, the editor would open far below the fold and it would look exactly like nothing happened. Measured actual scroll position before and after the click, with and without reduced-motion: WebKit scrolled 0 → 1319px and put the editor at the top of the viewport. Chromium scrolled 0 → 1320px. The engines agree.

  4. Not a Safari version floor. I pulled all 36 shipped production JS chunks and re-emitted each through esbuild targeting Safari 13 through 18. 24 of 36 need rewriting for Safari 13; zero need rewriting at Safari 14 or above. A parse error is the only failure that kills every island at once, and no Safari from 14 onward can hit one here. The only two runtime APIs above that floor are Array.prototype.at(-n) and crypto.randomUUID(), both shipped in Safari 15.4 (March 2022), and neither is reachable from the two buttons you named.

So I'm closing this as not reproducible — not as "not real". Something happened to you, and the most likely candidates left are all on your side of the wire: a Safari older than 15.4, Lockdown Mode, or a content blocker.

Please reopen if you can still hit it, and if you can, these five answers would crack it fast:

  1. Exact Safari version and OS.
  2. Does it still happen today? The profile page was restructured by #1503 since you filed.
  3. "Nothing in the dev console" — no errors, or no network request? Those point at opposite halves of the stack.
  4. Lockdown Mode or a content blocker active?
  5. For the Lab half: does "+ New" fail to open the form, or does the form open and Create do nothing? Those are unrelated code paths, and whether this is really "two surfaces, one cause" depends on the answer.

One thing your report did surface that I'm keeping: the Edit button is an island whose only job is to shout at a different island 600 lines further down the page, and it gives no feedback at all if nobody is listening. That's a genuinely bad shape — any silent failure in the second island looks precisely like your bug — and it's worth collapsing on its own merits. Filed separately as #1657 rather than smuggled into this one.

Sign in to post a ripple.