About unexpected char sequences into binary data

[saoirse-iontach]

Binary data are wrapped into <!-- ... --> comments.
But data may contains a --> char sequence!
Then we should reopen comment, but the parser doesn't allow that.

Instead of comment, we could wrap into script element,
Then handle closing with a custom src script.

<script>function onclose(){
  // if (true) /* then ends on EOF with a noop document write */
  // if (count--) /* pre recorded value */
  // if (!...endsWith('${win_1252_data.slice(-42)}')) /* pre recorded value */
  if (!document.currentScript.textContent.endsWith('===EOD===')) /* unseen custom marker */
  {
    document.write('<script src="data:,onclose()">');
  }
}</script>

<script src="data:,onclose()">Wrapped binary data....
...with unexpected</script>and continuation...
...with another unexpected</script probably illformed...>and...
...and finally ===EOD===</script>

Note that we have to restore original </script...> when joining all script text contents.

---

So in build process:

• If data doesn't contains -->, then use comment
• else if data doesn't contains </script, then use (no executable) script element
• else do complex build:
  • record all </script...> end tag
  • use script element with close handler
  • use appropriated post-processing to join all script text contents and recorded end tags.

[gildas-lormeau]

Actually, this is already implemented in SingleFile, here are the list of alternate tags which can be used when the binary data contains the closing tag: https://github.com/gildas-lormeau/single-file-core/blob/main/processors/compression/compression.js#L44-L49.

You can see it in action in the presentation itself (https://github.com/gildas-lormeau/Polyglot-HTML-ZIP-PNG/raw/main/en-EN/dist/presentation-polyglot-png-zip-html_en-EN.html) which uses good old <xmp> and </xmp> tags for the ZIP data.

[saoirse-iontach]

Effectively, usable tags for RawText (and Script Data) are:

• script and style (with unknow type)
• noscript, noframes, noembed,
• xpm
• iframe
• plaintext (witch runs to EOF!)

About plaintext, there is no end tag. It may be very interesting!

About end tag rexexp, they should include \b[^>]*   _{(=> /<\/tag\b[^>]*>/i )}

Thank you for all your help :-)

[gildas-lormeau]

Thank you for the info! I'll try to add the missing tags in SingleFile.