Crypto: Duplicate message index after using the share extension

[thegcat]

Steps to reproduce

1. Use Element iOS
2. Go to another App
3. Take a screenshot
4. Post screenshot to Element iOS using the share sheet
5. Write another message in the room the screenshot was posted to before anyone else could do anything (even the typing indicator from another user seems to remove this problem)

Outcome

What did you expect?

Being able to post a picture to a room using the iOS share sheet and then post a message that can be decrypted by other participants and my other devices in that chat.

What happened instead?

{
  "type": "m.room.message",
  "content": {
    "msgtype": "m.bad.encrypted",
    "body": "** Unable to decrypt: DecryptionError: Duplicate message index, possible replay attack: xxx **"
  }
}

Your phone model

iPhone 14 Pro

Operating system version

iOS 16.3.1

Application version

Element 1.10.2 (20230210134131)

Homeserver

{"server":{"name":"Synapse","version":"1.77.0"}}

Will you send logs?

No

[thegcat]

This is the same defect as #4104, not sure if this counts as a duplicate or not.

[Velin92]

Hi thanks for the report, let's keep this issue since you're report is also more descriptive.
Could you also provide a screen recording of the issue if possible?
Since I am not able to reproduce it.
Thanks.

[thegcat]

https://share.icloud.com/photos/0d8CTVORQIF127jLK0hjFnkBg

[thegcat]

This is the view from someone else in the room on Element Desktop Linux (Element version: 1.11.23 Olm version: 3.2.12). This is also what I see on Element Desktop macOS (Version von Element: 1.11.23 Version von Olm: 3.2.12). Interestingly enough the other person can decrypt the message on Element Android.

[Velin92]

Uhm I am not having any troubles on El-iOS, and since it's handled correctly on El-Android too, this might actually be an El-Web issue.
Maybe would be better to report this issue to the El-Web repo (since the desktop versions is essentially the same as the web version) and link it here @thegcat

[Velin92]

https://github.com/vector-im/element-web @thegcat here is the repo

[thegcat]

I thought this issue also existed between Element iOS, i.e. doing this on my iPhone would show the issue on my iPad, but I just checked this with the example I posted above and the message can be decrypted on the iPad.

[Velin92]

It's then very likely an el-web issue then closing this one here, but I suggest opening an issue there and linking this.
Thanks for the report!

[thegcat]

Uhm I am not having any troubles on El-iOS, and since it's handled correctly on El-Android too, this might actually be an El-Web issue.

I'm not quite sure this is only an element Web issue.

If I make a screenshot in element and send it via the share sheet while element iOS is still the currently running app behind the share sheet I cannot reproduce this issue. I.e. I would not say this is only an element Web issue, as the state of the element iOS is a factor.

[thegcat]

It's then very likely an el-web issue then closing this one here, but I suggest opening an issue there and linking this.

See above, the state of element iOS is relevant to reproduce this bug, and I don't think it's an element Web only bug. I would go so far as suspect Web to work correctly and Android and iOS not checking this sort of decryption issues (re-using a message index), but I am not knowledgeable enough wrt the encryption to say which way it is.

As a side-note, I have to say that being sent around within the Element eco-system to bugs that seem to already have been opened, fixed, but still creep up, and then in effect setting a +1 on a bug that seems to be abandoned for 2 years, doesn't offer neither a unified nor a generally positive impression of the project. Presenting "Element" as a unified branding to users and then telling them to go to that other element thing might not be easily understable and an experience inviting more bug reports to average users.

[Velin92]

Hello again! @thegcat
If you check the "view souce" option of a message you can check the event id of the message itself, to see if the same index/same message identifiers are being used.
In trying to reproduce the issue I see that they have two different IDs so it's definitely not this issue.
Also i tried to reproduce the bug with other non element clients, and it does not happen.
As I said is very likely a El-Web issue.
Thanks.

[richvdh]

@thegcat may be seeing #7499