renderTweetTextJSX now accepts skipSpans=[] as third param.
Skip-span boundaries are added to the pts split set so a trailing
media t.co inside a plain segment still gets isolated and suppressed—
not re-linked by linkifyText. Early return only when both anns and
skipSpans are empty.
TweetCard computes mediaSkipSpans after building photos/videoItems:
for each rawMedia item whose src resolved (photo src match; any video),
resolveEntityBounds(m, ft, m.url) gives the precise [s,e] span using
indices/fromIndex first, indexOf fallback—then the span is passed to
renderTweetTextJSX so the t.co attachment URL is silently dropped.
Server (routes.rs):
POST /api/util/resolve-tco — unauthenticated, accepts JSON array of
https://t.co/<alphanumeric> URLs only (strict regex validation, no SSRF
via input), capped at 50 per batch, 3 s timeout, redirect(Policy::none)
so the server only ever touches t.co itself. HEAD first, GET fallback if
HEAD returns no Location. Location sanitized to http/https only —
javascript:/data:/etc. fall back to the original t.co.
api.js:
resolveTcoUrls(urls) — project-convention wrapper for the new endpoint.
Returns {} on failure (callers degrade gracefully to bare t.co links).
TweetPreview.jsx:
After tweet data loads, per-tweet range-based coverage detection:
builds covered [start,end) from existing entity fromIndex/toIndex or
indices fallback, then finds regex matches whose span is NOT covered.
Resolves unique uncovered t.co URLs via resolveTcoUrls(), synthesises
one entity per occurrence (with exact fromIndex/toIndex so normalizeUrlAnn
gets correct bounds even for duplicate t.co URLs in the same tweet).
Augments entities.urls before setTweets() so all rendering paths
see expanded URLs.
