mirror of
https://github.com/hanzoai/mail.git
synced 2026-09-23 23:53:02 +00:00
Fork of github.com/mjl-/mox at 5656ca6d1b
(upstream v0.0.15 plus 55 commits).
Rename the module path in go.mod and the import paths in 268 Go files. Also
update the five scripts that "go install" the module (genwebsite.sh,
docker-release.sh, apidiff.sh, test-upgrade.sh, Dockerfile.release), which
would otherwise build upstream rather than this fork.
Add NOTICE recording the upstream project, author, license, and fork point,
and stating which license covers which files: MIT for the mox source and our
changes, MPL-2.0 for publicsuffix/public_suffix_list.txt alone, and the
vendored dependencies' own licenses under licenses/. Embed NOTICE so
"mail licenses" prints the lineage from the running binary. LICENSE.MIT,
LICENSE.MPLv2.0 and the licenses/ tree are preserved verbatim; no code is
relicensed.
The mox command name, config keys, data directory layout and website content
are unchanged. LLM.md records the fork point, the rename, what was left alone,
and the steps to stand the server up.
Co-authored-by: Hanzo Dev <[email protected]>
345 lines
11 KiB
Go
345 lines
11 KiB
Go
package webmail
|
|
|
|
import (
|
|
"errors"
|
|
"fmt"
|
|
"io"
|
|
"log/slog"
|
|
"mime"
|
|
"net/textproto"
|
|
"net/url"
|
|
"strings"
|
|
|
|
"golang.org/x/text/encoding/ianaindex"
|
|
|
|
"github.com/hanzoai/mail/dns"
|
|
"github.com/hanzoai/mail/message"
|
|
"github.com/hanzoai/mail/mlog"
|
|
"github.com/hanzoai/mail/mox-"
|
|
"github.com/hanzoai/mail/moxio"
|
|
"github.com/hanzoai/mail/smtp"
|
|
"github.com/hanzoai/mail/store"
|
|
"slices"
|
|
)
|
|
|
|
// todo: we should have all needed information for messageItem in store.Message (perhaps some data in message.Part) for fast access, not having to parse the on-disk message file.
|
|
|
|
var wordDecoder = mime.WordDecoder{
|
|
CharsetReader: func(charset string, r io.Reader) (io.Reader, error) {
|
|
switch strings.ToLower(charset) {
|
|
case "", "us-ascii", "utf-8":
|
|
return r, nil
|
|
}
|
|
enc, _ := ianaindex.MIME.Encoding(charset)
|
|
if enc == nil {
|
|
enc, _ = ianaindex.IANA.Encoding(charset)
|
|
}
|
|
if enc == nil {
|
|
return r, fmt.Errorf("unknown charset %q", charset)
|
|
}
|
|
return enc.NewDecoder().Reader(r), nil
|
|
},
|
|
}
|
|
|
|
// Attempt q/b-word-decode name, coming from Content-Type "name" field or
|
|
// Content-Disposition "filename" field.
|
|
//
|
|
// RFC 2231 specify an encoding for non-ascii values in mime header parameters. But
|
|
// it appears common practice to instead just q/b-word encode the values.
|
|
// Thunderbird and gmail.com do this for the Content-Type "name" parameter.
|
|
// gmail.com also does that for the Content-Disposition "filename" parameter, where
|
|
// Thunderbird uses the RFC 2231-defined encoding. Go's mime.ParseMediaType parses
|
|
// the mechanism specified in RFC 2231 only. The value for "name" we get here would
|
|
// already be decoded properly for standards-compliant headers, like
|
|
// "filename*0*=UTF-8”%...; filename*1*=%.... We'll look for Q/B-word encoding
|
|
// markers ("=?"-prefix or "?="-suffix) and try to decode if present. This would
|
|
// only cause trouble for filenames having this prefix/suffix.
|
|
func tryDecodeParam(log mlog.Log, name string) string {
|
|
if name == "" || !strings.HasPrefix(name, "=?") && !strings.HasSuffix(name, "?=") {
|
|
return name
|
|
}
|
|
// todo: find where this is allowed. it seems quite common. perhaps we should remove the pedantic check?
|
|
if mox.Pedantic {
|
|
log.Debug("attachment contains rfc2047 q/b-word-encoded mime parameter instead of rfc2231-encoded", slog.String("name", name))
|
|
return name
|
|
}
|
|
s, err := wordDecoder.DecodeHeader(name)
|
|
if err != nil {
|
|
log.Debugx("q/b-word decoding mime parameter", err, slog.String("name", name))
|
|
return name
|
|
}
|
|
return s
|
|
}
|
|
|
|
// todo: mime.FormatMediaType does not wrap long lines. should do it ourselves, and split header into several parts (if commonly supported).
|
|
|
|
func messageItemMoreHeaders(moreHeaders []string, pm ParsedMessage) (l [][2]string) {
|
|
for _, k := range moreHeaders {
|
|
k = textproto.CanonicalMIMEHeaderKey(k)
|
|
for _, v := range pm.Headers[k] {
|
|
l = append(l, [2]string{k, v})
|
|
}
|
|
}
|
|
return l
|
|
}
|
|
|
|
func messageItem(log mlog.Log, m store.Message, state *msgState, moreHeaders []string) (MessageItem, error) {
|
|
headers := len(moreHeaders) > 0
|
|
pm, err := parsedMessage(log, &m, state, false, true, headers)
|
|
if err != nil && errors.Is(err, message.ErrHeader) && headers {
|
|
log.Debugx("load message item without parsing headers after error", err, slog.Int64("msgid", m.ID))
|
|
pm, err = parsedMessage(log, &m, state, false, true, false)
|
|
}
|
|
if err != nil {
|
|
return MessageItem{}, fmt.Errorf("parsing message %d for item: %v", m.ID, err)
|
|
}
|
|
// Clear largish unused data.
|
|
m.MsgPrefix = nil
|
|
m.ParsedBuf = nil
|
|
l := messageItemMoreHeaders(moreHeaders, pm)
|
|
return MessageItem{m, pm.envelope, pm.attachments, pm.isSigned, pm.isEncrypted, true, l}, nil
|
|
}
|
|
|
|
func parsedMessage(log mlog.Log, m *store.Message, state *msgState, full, msgitem, msgitemHeaders bool) (pm ParsedMessage, rerr error) {
|
|
pm.ViewMode = store.ModeText // Valid default, in case this makes it to frontend.
|
|
|
|
if full || msgitem || state.newPreviews != nil && m.Preview == nil {
|
|
if !state.ensurePart(*m, true) {
|
|
return pm, state.err
|
|
}
|
|
if full {
|
|
pm.Part = *state.part
|
|
}
|
|
} else {
|
|
if !state.ensurePart(*m, false) {
|
|
return pm, state.err
|
|
}
|
|
}
|
|
if state.newPreviews != nil && m.Preview == nil {
|
|
s, err := state.part.Preview(log)
|
|
if err != nil {
|
|
log.Infox("generating preview", err, slog.Int64("msgid", m.ID))
|
|
}
|
|
// Set preview on m now, and let it be saved later on.
|
|
m.Preview = &s
|
|
state.newPreviews[m.ID] = s
|
|
}
|
|
|
|
// todo: we should store this form in message.Part, requires a data structure update.
|
|
|
|
convertAddrs := func(l []message.Address) []MessageAddress {
|
|
r := make([]MessageAddress, len(l))
|
|
for i, a := range l {
|
|
d, err := dns.ParseDomain(a.Host)
|
|
log.Check(err, "parsing domain")
|
|
if err != nil {
|
|
d = dns.Domain{ASCII: a.Host}
|
|
}
|
|
r[i] = MessageAddress{a.Name, a.User, d}
|
|
}
|
|
return r
|
|
}
|
|
|
|
if full || msgitem {
|
|
env := MessageEnvelope{}
|
|
if state.part.Envelope != nil {
|
|
e := *state.part.Envelope
|
|
env.Date = e.Date
|
|
env.Subject = e.Subject
|
|
env.InReplyTo = e.InReplyTo
|
|
env.MessageID = e.MessageID
|
|
env.From = convertAddrs(e.From)
|
|
env.Sender = convertAddrs(e.Sender)
|
|
env.ReplyTo = convertAddrs(e.ReplyTo)
|
|
env.To = convertAddrs(e.To)
|
|
env.CC = convertAddrs(e.CC)
|
|
env.BCC = convertAddrs(e.BCC)
|
|
}
|
|
pm.envelope = env
|
|
}
|
|
|
|
if (full || msgitemHeaders) && state.part.BodyOffset > 0 {
|
|
hdrs, err := state.part.Header()
|
|
if err != nil {
|
|
return ParsedMessage{}, fmt.Errorf("parsing headers: %w", err)
|
|
}
|
|
pm.Headers = hdrs
|
|
|
|
pm.ListReplyAddress = parseListPostAddress(hdrs.Get("List-Post"))
|
|
} else {
|
|
pm.Headers = map[string][]string{}
|
|
}
|
|
|
|
pm.Texts = []string{}
|
|
|
|
// We track attachments from multipart/mixed differently from other attachments.
|
|
// The others are often inline, sometimes just some logo's in HTML alternative
|
|
// messages. We want to have our mixed attachments at the start of the list, but
|
|
// our descent-first parsing would result in inline messages first in the typical
|
|
// message.
|
|
var attachmentsMixed []Attachment
|
|
var attachmentsOther []Attachment
|
|
|
|
addAttachment := func(a Attachment, isMixed bool) {
|
|
if isMixed {
|
|
attachmentsMixed = append(attachmentsMixed, a)
|
|
} else {
|
|
attachmentsOther = append(attachmentsOther, a)
|
|
}
|
|
}
|
|
|
|
// todo: how should we handle messages where a user prefers html, and we want to show it, but it's a DSN that also has textual-only parts? e.g. gmail's dsn where the first part is multipart/related with multipart/alternative, and second part is the regular message/delivery-status. we want to display both the html and the text.
|
|
|
|
var usePart func(p message.Part, index int, parent *message.Part, path []int, parentMixed bool)
|
|
usePart = func(p message.Part, index int, parent *message.Part, path []int, parentMixed bool) {
|
|
mt := p.MediaType + "/" + p.MediaSubType
|
|
newParentMixed := mt == "MULTIPART/MIXED"
|
|
for i, sp := range p.Parts {
|
|
if mt == "MULTIPART/SIGNED" && i >= 1 {
|
|
continue
|
|
}
|
|
usePart(sp, i, &p, append(slices.Clone(path), i), newParentMixed)
|
|
}
|
|
switch mt {
|
|
case "TEXT/PLAIN", "/":
|
|
// Don't include if Content-Disposition attachment.
|
|
if full || msgitem {
|
|
disp, name, err := p.DispositionFilename()
|
|
if err != nil && errors.Is(err, message.ErrParamEncoding) {
|
|
log.Debugx("parsing disposition/filename", err)
|
|
} else if err != nil {
|
|
rerr = fmt.Errorf("reading disposition/filename: %v", err)
|
|
return
|
|
}
|
|
if strings.EqualFold(disp, "attachment") {
|
|
addAttachment(Attachment{path, name, p}, parentMixed)
|
|
return
|
|
}
|
|
}
|
|
|
|
if full {
|
|
buf, err := io.ReadAll(&moxio.LimitReader{R: p.ReaderUTF8OrBinary(), Limit: 2 * 1024 * 1024})
|
|
if err != nil {
|
|
rerr = fmt.Errorf("reading text part: %v", err)
|
|
return
|
|
}
|
|
pm.Texts = append(pm.Texts, string(buf))
|
|
pm.TextPaths = append(pm.TextPaths, slices.Clone(path))
|
|
}
|
|
|
|
case "TEXT/HTML":
|
|
pm.HasHTML = true
|
|
if full && pm.HTMLPath == nil {
|
|
pm.HTMLPath = slices.Clone(path)
|
|
}
|
|
|
|
default:
|
|
// todo: see if there is a common nesting messages that are both signed and encrypted.
|
|
if parent == nil && mt == "MULTIPART/SIGNED" {
|
|
pm.isSigned = true
|
|
}
|
|
if parent == nil && mt == "MULTIPART/ENCRYPTED" {
|
|
pm.isEncrypted = true
|
|
}
|
|
// todo: possibly do not include anything below multipart/alternative that starts with text/html, they may be cids. perhaps have a separate list of attachments for the text vs html version?
|
|
if p.MediaType != "MULTIPART" {
|
|
var parentct string
|
|
if parent != nil {
|
|
parentct = parent.MediaType + "/" + parent.MediaSubType
|
|
}
|
|
|
|
// Recognize DSNs.
|
|
if parentct == "MULTIPART/REPORT" && index == 1 && (mt == "MESSAGE/GLOBAL-DELIVERY-STATUS" || mt == "MESSAGE/DELIVERY-STATUS") {
|
|
if full {
|
|
buf, err := io.ReadAll(&moxio.LimitReader{R: p.ReaderUTF8OrBinary(), Limit: 1024 * 1024})
|
|
if err != nil {
|
|
rerr = fmt.Errorf("reading text part: %v", err)
|
|
return
|
|
}
|
|
pm.Texts = append(pm.Texts, string(buf))
|
|
pm.TextPaths = append(pm.TextPaths, slices.Clone(path))
|
|
}
|
|
return
|
|
}
|
|
if parentct == "MULTIPART/REPORT" && index == 2 && (mt == "MESSAGE/GLOBAL-HEADERS" || mt == "TEXT/RFC822-HEADERS") {
|
|
if full {
|
|
buf, err := io.ReadAll(&moxio.LimitReader{R: p.ReaderUTF8OrBinary(), Limit: 1024 * 1024})
|
|
if err != nil {
|
|
rerr = fmt.Errorf("reading text part: %v", err)
|
|
return
|
|
}
|
|
pm.Texts = append(pm.Texts, string(buf))
|
|
pm.TextPaths = append(pm.TextPaths, slices.Clone(path))
|
|
}
|
|
return
|
|
}
|
|
if parentct == "MULTIPART/REPORT" && index == 2 && (mt == "MESSAGE/GLOBAL" || mt == "TEXT/RFC822") {
|
|
addAttachment(Attachment{path, "original.eml", p}, parentMixed)
|
|
return
|
|
}
|
|
|
|
name := tryDecodeParam(log, p.ContentTypeParams["name"])
|
|
if name == "" && (full || msgitem) {
|
|
// todo: should have this, and perhaps all content-* headers, preparsed in message.Part?
|
|
h, err := p.Header()
|
|
log.Check(err, "parsing attachment headers", slog.Int64("msgid", m.ID))
|
|
cp := h.Get("Content-Disposition")
|
|
if cp != "" {
|
|
_, params, err := mime.ParseMediaType(cp)
|
|
log.Check(err, "parsing content-disposition", slog.String("cp", cp))
|
|
name = tryDecodeParam(log, params["filename"])
|
|
}
|
|
}
|
|
addAttachment(Attachment{path, name, p}, parentMixed)
|
|
}
|
|
}
|
|
}
|
|
usePart(*state.part, -1, nil, []int{}, false)
|
|
|
|
pm.attachments = []Attachment{}
|
|
pm.attachments = append(pm.attachments, attachmentsMixed...)
|
|
pm.attachments = append(pm.attachments, attachmentsOther...)
|
|
|
|
if rerr == nil {
|
|
pm.ID = m.ID
|
|
}
|
|
return
|
|
}
|
|
|
|
// parses List-Post header, returning an address if it could be found, and nil otherwise.
|
|
func parseListPostAddress(s string) *MessageAddress {
|
|
/*
|
|
Examples:
|
|
List-Post: <mailto:[email protected]>
|
|
List-Post: <mailto:[email protected]> (Postings are Moderated)
|
|
List-Post: <mailto:[email protected]?subject=list%20posting>
|
|
List-Post: NO (posting not allowed on this list)
|
|
List-Post: <https://groups.google.com/group/golang-dev/post>, <mailto:[email protected]>
|
|
*/
|
|
s = strings.TrimSpace(s)
|
|
for s != "" {
|
|
if !strings.HasPrefix(s, "<") {
|
|
return nil
|
|
}
|
|
addr, ns, found := strings.Cut(s[1:], ">")
|
|
if !found {
|
|
return nil
|
|
}
|
|
if strings.HasPrefix(addr, "mailto:") {
|
|
u, err := url.Parse(addr)
|
|
if err != nil {
|
|
return nil
|
|
}
|
|
addr, err := smtp.ParseAddress(u.Opaque)
|
|
if err != nil {
|
|
return nil
|
|
}
|
|
return &MessageAddress{User: addr.Localpart.String(), Domain: addr.Domain}
|
|
}
|
|
s = strings.TrimSpace(ns)
|
|
s = strings.TrimPrefix(s, ",")
|
|
s = strings.TrimSpace(s)
|
|
}
|
|
return nil
|
|
}
|