Fix inbound-email data loss, MIME truncation, and header injection

- server/data.js: preserve source/htmlBody/inboundTo/messageId across
  server restarts (sanitizeLoadedContactSubmissions was silently
  dropping them on reload from disk)
- cloudflare/email-worker.js: rewrite MIME parsing to split on the
  actual boundary marker instead of any literal "--", unfold
  multi-line headers, and correctly recombine multi-byte UTF-8 in
  quoted-printable decoding
- server/routes/inbound-email.js: validate Message-ID against RFC 5322
  grammar before storing/using it, and compare the webhook secret with
  timingSafeEqual to match the rest of the codebase's auth checks
- server/routes/contact.js: re-validate messageId at the point it's
  injected into outgoing In-Reply-To/References headers; move the
  allowed reply-from addresses into a shared config constant
- src/AdminPage.tsx: 30s inbox poll now syncs field updates (e.g.
  archived) on already-loaded submissions instead of only appending
  new ones; consolidate the duplicated from-address list
- .claude/launch.json: add a vite dev server preview config used to
  verify these changes

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This commit is contained in:
nmemmert
2026-07-02 08:21:50 -04:00
parent fbb9a24f32
commit 17c9cbbc8b
7 changed files with 147 additions and 52 deletions
+100 -35
View File
@@ -60,23 +60,7 @@ async function streamToText(stream) {
}
function parseEmail(raw, message) {
const lines = raw.split(/\r?\n/)
// Parse headers (everything before the first blank line)
const headers = {}
let bodyStart = 0
for (let i = 0; i < lines.length; i++) {
if (lines[i].trim() === '') {
bodyStart = i + 1
break
}
const colon = lines[i].indexOf(':')
if (colon > 0) {
const key = lines[i].slice(0, colon).trim().toLowerCase()
const val = lines[i].slice(colon + 1).trim()
if (!headers[key]) headers[key] = val
}
}
const { headers, body } = splitHeadersAndBody(raw)
const subject = decodeHeaderValue(headers['subject'] ?? '(no subject)')
const from = message.from ?? headers['from'] ?? ''
@@ -84,9 +68,7 @@ function parseEmail(raw, message) {
const date = headers['date'] ?? new Date().toISOString()
const messageId = headers['message-id'] ?? ''
const bodyLines = lines.slice(bodyStart)
const plainText = extractPart(raw, bodyLines, 'text/plain')
const htmlBody = extractPart(raw, bodyLines, 'text/html')
const { plainText, htmlBody } = extractBodies(headers['content-type'] ?? '', body)
return {
from,
@@ -100,28 +82,111 @@ function parseEmail(raw, message) {
}
}
function extractPart(raw, bodyLines, contentType) {
const escaped = contentType.replace('/', '\\/')
const re = new RegExp(`Content-Type: ${escaped}[^\\r\\n]*\\r?\\n(?:[^\\r\\n]+\\r?\\n)*\\r?\\n([\\s\\S]*?)(?=--|$)`, 'i')
const match = re.exec(raw)
if (match) return decodeEmailBody(match[1]).trim()
// Splits an RFC 5322 message (or MIME part) into its unfolded header map and raw body string.
function splitHeadersAndBody(raw) {
const match = /\r?\n\r?\n/.exec(raw)
const headerBlock = match ? raw.slice(0, match.index) : raw
const body = match ? raw.slice(match.index + match[0].length) : ''
// RFC 2822 header folding: a line starting with SP/TAB continues the previous header line.
const unfolded = headerBlock.replace(/\r?\n[ \t]+/g, ' ')
const headers = {}
for (const line of unfolded.split(/\r?\n/)) {
const colon = line.indexOf(':')
if (colon <= 0) continue
const key = line.slice(0, colon).trim().toLowerCase()
const val = line.slice(colon + 1).trim()
if (!headers[key]) headers[key] = val
}
return { headers, body }
}
if (contentType === 'text/plain') {
const joined = bodyLines.join('\n')
if (/<[a-z][\s\S]*>/i.test(joined)) {
return joined.replace(/<[^>]+>/g, '').replace(/&nbsp;/g, ' ').replace(/&amp;/g, '&').replace(/\s{3,}/g, '\n\n').trim()
// Recursively walks a (possibly multipart) MIME body and returns the first text/plain and text/html parts found.
function extractBodies(topContentType, topBody) {
let plainText = null
let htmlBody = null
function visit(contentType, body, transferEncoding) {
const type = (contentType.split(';')[0] || 'text/plain').trim().toLowerCase()
if (type.startsWith('multipart/')) {
const boundary = getBoundary(contentType)
if (!boundary) return
for (const part of splitOnBoundary(body, boundary)) {
const { headers, body: partBody } = splitHeadersAndBody(part)
visit(headers['content-type'] ?? 'text/plain', partBody, headers['content-transfer-encoding'] ?? '')
}
return
}
return joined.trim()
const decoded = decodeBody(body, transferEncoding)
if (type === 'text/html' && htmlBody === null) htmlBody = decoded.trim()
if (type === 'text/plain' && plainText === null) plainText = decoded.trim()
}
return null
visit(topContentType || 'text/plain', topBody, '')
if (plainText === null && htmlBody !== null) {
plainText = htmlBody.replace(/<[^>]+>/g, '').replace(/&nbsp;/g, ' ').replace(/&amp;/g, '&').replace(/\s{3,}/g, '\n\n').trim()
}
if (plainText === null) plainText = ''
return { plainText, htmlBody }
}
function getBoundary(contentType) {
const match = /boundary\s*=\s*"([^"]+)"|boundary\s*=\s*([^;\s]+)/i.exec(contentType)
if (!match) return null
return match[1] ?? match[2]
}
// Splits a multipart body on its boundary markers, ignoring the preamble/epilogue.
function splitOnBoundary(body, boundary) {
const escaped = boundary.replace(/[.*+?^${}()|[\]\\]/g, '\\$&')
const re = new RegExp(`(?:^|\\r?\\n)--${escaped}(--)?(?:\\r?\\n|$)`, 'g')
const parts = []
let lastIndex = 0
let started = false
let match
while ((match = re.exec(body)) !== null) {
if (started) parts.push(body.slice(lastIndex, match.index))
started = true
lastIndex = match.index + match[0].length
if (match[1]) break // final boundary: "--boundary--"
}
return parts
}
function decodeBody(text, transferEncoding) {
const enc = (transferEncoding || '').trim().toLowerCase()
if (enc === 'base64') {
try {
const binary = atob(text.replace(/\s+/g, ''))
return new TextDecoder('utf-8').decode(Uint8Array.from(binary, c => c.charCodeAt(0)))
} catch {
return ''
}
}
if (enc === 'quoted-printable') {
return decodeEmailBody(text)
}
return text
}
function decodeEmailBody(text) {
// Handle quoted-printable encoding (=XX hex sequences and soft line breaks)
return text
.replace(/=\r?\n/g, '')
.replace(/=([0-9A-Fa-f]{2})/g, (_, hex) => String.fromCharCode(parseInt(hex, 16)))
// Handle quoted-printable encoding (=XX hex sequences and soft line breaks).
// Decode into raw bytes first, then run through TextDecoder so multi-byte
// UTF-8 sequences split across multiple =XX escapes recombine correctly.
const unfolded = text.replace(/=\r?\n/g, '')
const bytes = []
for (let i = 0; i < unfolded.length; i++) {
if (unfolded[i] === '=' && /^[0-9A-Fa-f]{2}$/.test(unfolded.slice(i + 1, i + 3))) {
bytes.push(parseInt(unfolded.slice(i + 1, i + 3), 16))
i += 2
} else {
bytes.push(unfolded.charCodeAt(i))
}
}
return new TextDecoder('utf-8').decode(Uint8Array.from(bytes))
}
function decodeHeaderValue(value) {