Skip to content

og: rename card to og-v2.png, social tags on one line each - #37

Merged
itsdangold merged 1 commit into
masterfrom
dang/og-slack
Sep 18, 2026
Merged

itsdangold merged 1 commit into
masterfrom
dang/og-slack

Conversation

@itsdangold

Copy link
Copy Markdown
Contributor

Slack still shows the unfurl without an image. Everything checkable from outside
is correct, so this targets the two things that could still explain it.

What I ruled out first

check result
og tags present, in <head> all five, plus image sub-properties
absolute image URL https://live.sentry.io/og-v2.png
32KB Slackbot limit page is 2990 bytes — 9% of it
publicly reachable 200 to Slackbot-LinkExpanding, no auth
robots.txt 404, nothing blocked
TLS chain full 4-cert chain, verifies clean
redirects none
image 200, image/png, 1200×630, fully opaque, 245KB
fleet 24/24 fresh requests returned 200

Facebook, LinkedIn and opengraph.xyz all resolve the image. Only Slack does not.

Change 1 — rename to og-v2.png

Slack caches the image separately from the page, keyed on the image URL, so
busting the page URL with ?v=2 never touched it. If Slack's proxy tried
og-v1.png while a region was still serving the old container during the
rollout, it cached that 404. A new filename is a guaranteed cache miss — which
is exactly what versioning the filename was for.

This is the one I expect to fix it.

Change 2 — one line per tag, block moved to the top of <head>

Prettier had wrapped og:image:alt and twitter:image:alt across three lines.
A real HTML parser does not care, which is why the validators pass — but
og:image:alt is a sub-property of og:image, and a line-oriented scraper that
trips on it can drop the whole image property with it. Cheap to rule out.

Also adds og:image:secure_url (redundant over https, harmless).

Verified

The smoke test asserted /og-v1.png, so it moves with the rename — otherwise
this would have gone red like last time. Built the image and ran the step:

all smoke-test assertions pass (including the renamed /og-v2.png)
og:image served: property="og:image" content="https://live.sentry.io/og-v2.png"
old path /og-v1.png now: 404  (expected 404)

After merging

Delete the old Slack messages and post the link fresh — an existing message
keeps its cached unfurl regardless of what the page now says.

…each

Every item on the standard Slack checklist already passed -- tags present and
absolute, page 2990 bytes against a 32KB limit, no robots.txt, publicly
reachable, full TLS chain, no redirects -- and Facebook, LinkedIn and
opengraph.xyz all resolve the image. Only Slack does not.

Two changes, for the two things that could still explain it.

Rename to og-v2.png. Slack caches the image separately from the page, keyed on
the image URL, so busting the page URL does not touch it. If its proxy tried
og-v1.png while a region was still serving the old container it would have
cached that 404. A new filename is a guaranteed miss, which is what versioning
it was for.

Put every social tag on a single line and move the block to the top of the
head. Prettier had wrapped og:image:alt and twitter:image:alt across three
lines. A proper HTML parser does not care, which is why the validators pass,
but og:image:alt is a sub-property of og:image, and a line-oriented scraper
that trips on it can drop the whole image property. Also adds
og:image:secure_url, which is redundant over https but harmless.

The smoke test asserted /og-v1.png, so it moves with the rename. Verified
against the built container: every assertion passes and the old path 404s.
@itsdangold
itsdangold merged commit 59b0bb4 into master Sep 18, 2026
10 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