Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix showing edited message for historical messages #714

Open
wants to merge 1 commit into
base: dev
Choose a base branch
from

Conversation

TobiasFella
Copy link
Member

Fixes #713

This is mostly just extracting the code from Room::Private::addNewMessageEvents in order to be able to reuse it in Room::Private::addHistoricalMessageEvents. Note that there is one change: the range of events being searched for the replaced event is changed from (events.begin(), it) to (events.begin(), events.end()) - Apparently this is required when handling historical messages; I don't know the exact reason, but don't see how it could cause problems.

@TobiasFella
Copy link
Member Author

CC @redstrate

Copy link

sonarqubecloud bot commented Nov 4, 2023

Kudos, SonarCloud Quality Gate passed!    Quality Gate passed

Bug A 0 Bugs
Vulnerability A 0 Vulnerabilities
Security Hotspot A 0 Security Hotspots
Code Smell A 1 Code Smell

66.7% 66.7% Coverage
0.0% 0.0% Duplication

@TobiasFella TobiasFella force-pushed the tobias/historical-edits branch from 0d504df to 604929f Compare November 5, 2023 15:39
@TobiasFella TobiasFella force-pushed the tobias/historical-edits branch from 604929f to b3795d9 Compare November 5, 2023 16:28
@KitsuneRal
Copy link
Member

KitsuneRal commented Nov 5, 2023

The reason why only a subrange is checked was simple optimisation: redacts and edits always come after the events they change. In case of historical messages it's meant to be the reverse: historical events from /messages are ordered newest to oldest, so the range to check would be something like (events.rbegin(), it) in reverse iterators or (it, events.end()) in forward iterators (with lookup going in different directions but that's not important in this case).
I don't see any issues with using the full range of events instead - those ranges are never so large.

return ep->id() == id;
});
if (targetIt != events.end())
*targetIt = makeReplaced(**targetIt, *msg);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: Since we're moving this block, can we use std::exchange. Ditto for the first branch too

@KitsuneRal
Copy link
Member

KitsuneRal commented Nov 5, 2023

Thinking further of the whole thing - what this doesn't cover is the case when an edit or a redaction applies to an event that resides in an even older, not yet received batch. I still wonder if #713 should be reported as the homeserver issue, rather. Doesn't mean I'm unwilling to cover up the problem on the client side, given that the issue already emerged. I would certainly add a warning in the logs for that situation, so that at least knowledgeable people would take heed.

@redstrate
Copy link
Contributor

Thinking further of the whole thing - what this doesn't cover is the case when an edit or a redaction applies to an event tier resides in an even older, not yet received batch.

I need to re-read this part of the spec again, but shouldn't that not matter? I would think if we the original message event, you must have the edits that came before it since you shouldn't be able to edit messages before they're sent. Maybe if the client is in a middle of a timeline, and they only see the message and some edits but not future ones. I'm not sure if that's possible in NeoChat/Quotient currently.

+1 to reporting this as a homeserver issue though, but we could at least do the best we can client-side.

Copy link
Member

@KitsuneRal KitsuneRal left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Revisiting this, and having re-read the spec again - I agree that aggregations (and de-aggregations, but that part is out of scope here for now) are hard (c) and both servers and clients have to do their best effort in tackling them. That being said, if we assume that aggregations have to be processed on the client side too then we have to do some extra stuff, namely two things:

  1. Make sure the library processes "orphaned" child events well (see below).
  2. Make sure the library processes aggregations in the right order - I'm not quite sure what will happen if, say, an event and its two replacements come in a single historical batch: I think the PR code as it is will display the older replacement instead of the newer one (bearing in mind that /messages?dir=b returns events in the reverse-chronological order; moveEventsToTimeline() takes care of re-reversing the order but processRedactionsAndEdits() apparently doesn't).

Comment on lines +2917 to +2923
else // FIXME: hide the replacing event when target arrives later
qCDebug(EVENTS)
<< "Replacing event" << msg->id()
<< "ignored: target event" << msg->replacedEvent()
<< "is not found";
// Same as with redactions above, the replaced event coming
// later will come already with the new content.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is applicable to events coming with sync but not with history. In case of historical events, I guess we should keep a map of "orphaned" redactions and replacements and resolve them as additional batches arrive. In absence of the parent event, historical redactions are probably a no-op while historical replacements should be shown as they are (and we can even mark them as edited).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Message edits lost after restarts
3 participants