Handle messages containing only end boundary, fixes #38

jhnstrk commented 5 months ago

Some browsers create empty messages when forms are posted - see #38 for examples.

The fix here adds a new state in the parser which is entered when the first hyphen after the first boundary is encountered. This is only the case for the very first boundary. Anything but a second hyphen after the first is considered an error.

tomchristie commented 5 months ago

Sounds feasible, let's get some clarity on it...

Is that spec compliant?
~~Can you provide an example (which browser and how to test)?~

Edit:

Apologies for the noise, I've not confirmed this but can see the following linked...

Firefox and Chrome produce empty messages like this, e.g. when POSTing a HTML form with only a unchecked checkbox.

(And... probably not relevant if it's spec compliant or not if that's the actual behavior.)

jhnstrk commented 5 months ago

To me the spec isn't 100% clear. rfc2046 appears to assume that there is always at least one body part, so this would not be valid. But it doesn't make it clear how a form with no parts should be communicated either.

Anyway, it's what Chrome (and probably Firefox, but I haven't tested myself) does now. Take the following text, adapted from the examples in #38

save as a test.html
open the file in Chrome
open Chrome dev tools
click the button on the form
inspect the payload sent to example.com (Network tab,--> Payload --> View Source)

<!DOCTYPE html>
<html>
  <head><meta charset="UTF-8"></head>
  <body>
        <form action="http://example.com/" enctype="multipart/form-data" method="post">
            <input type="checkbox" id="check" name="check">
            <label for="check">Breaks unless you check it</label>
            <input type="submit" value="Submit">
        </form>  
  </body>
</html>

I got the following payload: (Chrome Version 124.0.6367.61 (Official Build) (64-bit))

------WebKitFormBoundaryBBDliyOmePSMlFiJ--

Clicking (enabling) the check box in the form gives valid content with 1 part.

Abdelhadi92 commented 3 weeks ago

any updates on this?

Kludex / python-multipart

Handle messages containing only end boundary, fixes #38 #142