Commit graph

195 commits

Author SHA1 Message Date
Bruno Windels
bc69e49cfb WIP2 2020-03-31 10:13:25 +02:00
Bruno Windels
b6a5a02a33 WIP 2020-03-30 23:56:03 +02:00
Bruno Windels
e4563135bf look at prev_content when handling member events 2020-03-30 21:59:44 +02:00
Bruno Windels
c830f50325 catch errors while persisting a new message 2020-03-30 21:33:04 +02:00
Bruno Windels
234c260339 dont modify fragments in comparer until txn succeeds 2020-03-30 20:46:52 +02:00
Bruno Windels
73ea09f668 Handle sync ordering vs back-fill ordering (see synapse #7164) resulting in dupe events
Also extra robustness while filling a gap

ignore duplicate events from synapse instead considering them an
overlapping event with the adjacent fragment
2020-03-30 20:37:36 +02:00
Bruno Windels
8354c58c07 also look for remote echos based on event_id 2020-03-23 23:00:33 +01:00
Bruno Windels
f02641c808 look for transaction_id in /messages response to delete pending events 2020-03-22 00:07:37 +01:00
Bruno Windels
6d68ec1bac move fillGap to room 2020-03-21 23:40:40 +01:00
Bruno Windels
44760a9f82 Revert "add logging related to fragments"
This reverts commit 2f355cabef.
2020-03-21 10:47:35 +01:00
Bruno Windels
c4a5de20c0 update previous fragment in comparer when appending a new line fragment 2020-03-21 10:45:09 +01:00
Bruno Windels
2f355cabef add logging related to fragments
both when rebuilding and when sorting will be used
2020-03-20 20:23:07 +01:00
Bruno Windels
7e8cc7ae00 need to return the response here, not the request wrapper
we were reading back a remote id of undefined because of this,
so when for some reason we never receive the message down from sync,
the pending message keeps sending on every load. The server ignores
the send though, because the transaction id is already used, and it returns
the remote id of the event that was already sent the previous time, but
as we were not storing this remote id, we'd just try again and again.

not receiving it through sync could have happened when we were sending a bunch of events
and then receiving (this is how we encountered this bug, while trying to repro another) the
response, but not yet the sync for the message that got wedged. Then we typed stuff on another client
so we would get a limited response for that room, and boom, we would not get the remote echo of the
event that was already sent (but because of this bug we didn't store the remote id) but no echo received yet (when we remove the pending event),
so it gets wedged!
2020-03-17 00:11:50 +01:00
Bruno Windels
b69efc3425 fix errors when replaying 2020-03-17 00:07:54 +01:00
Bruno Windels
06e43d9ee3 fix all tests 2020-03-14 21:38:37 +01:00
Bruno Windels
56c87b3966 better naming 2020-03-14 20:49:15 +01:00
Bruno Windels
ac4a7f971b only apply sync changes in room summary once txn is committed
- also remove m.room.aliases support as they were wrongly implemented
   and now obsolete
 - don't count invited and joined members according to m.room.member
   events anymore as it was also wrongly implemented
   (only when prev!==new membership, but on initial sync we only get
   last member event, which might have been a nick change
2020-03-14 20:46:49 +01:00
Bruno Windels
fc741272ba only apply sync changes in session once txn is committed 2020-03-14 20:45:36 +01:00
Bruno Windels
224d56698a only set new live key when creating a fragment after the txn succeeds
when doing a limited sync, and a new fragment is created,
this._lastLiveKey is updated immediately. If the transaction
would then fail, the fragmentId in this._lastLiveKey was incremented
but the fragment wasn't written to the store, so if sync is resumed
and would subsequently succeed, fragmentIds would be assigned to events
that don't have a corresponding fragment in the timelineFragment store.

This would throw errors when trying to load the timeline,
breaking the whole app.

This changes SyncWriter to only update this._lastLiveKey in
the emit phase, when the transactions has been committed already.
2020-01-04 20:06:49 +01:00
Bruno Windels
bdc2c3d9ad cleanup: storage is not used in SyncWriter
as the transaction is now always passed as an argument, it never
creates one on its own.
2020-01-04 20:04:57 +01:00
Bruno Windels
1d9a5c490a add record/replay network layer 2019-12-23 14:28:42 +01:00
Bruno Windels
aa86748cdd pull fetch code out of homeserverapi 2019-12-23 14:28:27 +01:00
Bruno Windels
35c090dcb4 add proper import/export functionality 2019-12-14 18:29:35 +01:00
Bruno Windels
8a55778c32 make sure aliases and heroes aren't empty when picking name 2019-11-22 09:29:18 +01:00
Bruno Windels
325077bcb0 try more domexception fixes 2019-11-21 18:28:18 +01:00
Bruno Windels
5ecf0a6717 fix error path on transaction abort error 2019-11-21 18:23:48 +01:00
Bruno Windels
ad7a150aaa use aliases and heroes to fall back room name to 2019-10-13 07:48:33 +02:00
Bruno Windels
b16f21867c preserve value for storage errors on add and put 2019-10-12 22:19:16 +02:00
Bruno Windels
201b70ee4f fix failing to sync new rooms 2019-10-12 22:18:36 +02:00
Bruno Windels
ac53b9099b only log error in sync loop 2019-10-12 22:18:19 +02:00
Bruno Windels
2fa5fa7890 add delete button in session picker 2019-10-12 21:16:48 +02:00
Bruno Windels
d367037332 very basic support for lazy loading 2019-10-12 20:24:09 +02:00
Bruno Windels
8616d79aad make fetch errors more informative by including method and url 2019-09-15 14:31:50 +02:00
Bruno Windels
2b2b4232e9 moar send logging 2019-09-15 12:25:14 +02:00
Bruno Windels
6f527fd2e0 use openCursor where openKeyCursor is not supported 2019-09-15 12:24:46 +02:00
Bruno Windels
bbf6943455 use .get fallback where .getKey is not supported (Edge 15) 2019-09-15 12:24:27 +02:00
Bruno Windels
8e590fe53b rename and use StorageError to deal with idb DOMerror 2019-09-15 12:23:54 +02:00
Bruno Windels
5a90b4673b fix CORS requests failing 2019-09-15 12:23:08 +02:00
Bruno Windels
09b9eff7c1 fix remaining errors to make login work 2019-09-08 10:19:16 +02:00
Bruno Windels
49a577991b login and session picker view models, sessions store 2019-07-31 00:06:53 +02:00
Bruno Windels
45cd85ead1 fixup: check for event entry with eventType, event isn't exposed anymore 2019-07-29 19:58:35 +02:00
Bruno Windels
8665bcb897 concat synced events in timeline with pending events for local echo 2019-07-29 19:53:58 +02:00
Bruno Windels
88a7d64091 fixup: delay not awaited, but not needed for now 2019-07-29 19:53:30 +02:00
Bruno Windels
9b94c4bb61 don't expose raw event object from entry, pending event doesn't have it
it only has content and *some* of the meta fields,
but we want to threat pendingevententry and evententry as one
and the same in the rest of the application, so don't give access
to entire event object.
2019-07-29 10:27:12 +02:00
Bruno Windels
56cee450d1 return syncing user id from PendingEventEntry.sender
add User class where we later can track display name, avatar, ...
2019-07-29 10:24:34 +02:00
Bruno Windels
b26f7df689 pending events are not certain to be sorted here 2019-07-29 09:54:34 +02:00
Bruno Windels
3ed72df620 put everything together to make it roughly work
no local echo yet, and send errors are being swallowed
2019-07-27 10:51:51 +02:00
Bruno Windels
851100b88a send unsent messages after first sync 2019-07-26 22:40:39 +02:00
Bruno Windels
707988f806 better handle remote echos and hookup in session and room 2019-07-26 22:33:33 +02:00
Bruno Windels
ccb722d766 more WIP 2019-07-26 22:03:57 +02:00
Bruno Windels
f3d1128f28 WIP 2019-07-01 10:00:29 +02:00
Bruno Windels
c5b2d0c8b2 WIP 2019-06-28 00:52:54 +02:00
Bruno Windels
cd9f25ea80 convert tabs to spaces where needed 2019-06-26 22:31:36 +02:00
Bruno Windels
90a38078e0 unused import 2019-06-26 22:19:40 +02:00
Bruno Windels
b882e0ef8a respect argument count of idb calls
lumia doesn't like undefined arguments if they
are being left out, so call the idb calls with
the exact amount of arguments
2019-06-26 22:02:00 +02:00
Bruno Windels
bbb5e35bcb wrap everything that can throw a idb DOMException in StorageError
as lumia gives very cryptic errors without a stacktrace.
2019-06-26 22:00:50 +02:00
Bruno Windels
0fd52be710 encode idb array keys as sortable strings
that's why numeric parts of the keys have to be encoded
as a fixed length, "big-endian" ordered strings, so
string sorting will also sort the numeric keys correctly.

this also assumes room ids don't contain the "|" character,
we should probably escape the separator at some point.
2019-06-26 21:55:33 +02:00
Bruno Windels
106146660c define storage keys to be 32bit for idb / web platform
as a preparation to serialize the eventIndex and fragmentId
as a 8 character string, part of a concatenated string PK,
as lumia doesn't support array keys.
2019-06-26 21:52:41 +02:00
Bruno Windels
ca4361248f fallback for platforms missing AbortController 2019-06-26 19:49:49 +02:00
Bruno Windels
9cd5257959 load more events when scrolled to top 2019-06-16 17:29:33 +02:00
Bruno Windels
b47898e025 show full stack on sync error 2019-06-16 15:49:00 +02:00
Bruno Windels
1917a528c7 replace ad hoc buttons and label with SyncStatusBar 2019-06-16 10:54:16 +02:00
Bruno Windels
a4bc2dd2b0 support isOwn on messages 2019-06-16 10:53:23 +02:00
Bruno Windels
0524f06722 remove logging 2019-06-03 00:31:21 +02:00
Bruno Windels
6bdf44d114 adjust comment 2019-06-03 00:30:37 +02:00
Bruno Windels
45528580ed fix comparator/comparer mismatch 2019-06-03 00:30:16 +02:00
Bruno Windels
3dbf5e727d process in incoming order (reverse-chronological order if backward)
makes code simpler, don't need fix to undo reverse ordering of
nonOverlappingEvents. reverse looking is very likely premature
optimization as well.
2019-06-03 00:18:52 +02:00
Bruno Windels
0407829b26 fix filling gaps with overlapping events
although event order remains wrong, as events are reversed.
step before removing premature optimization, so it's in the git
commit log
2019-06-03 00:11:29 +02:00
Bruno Windels
c9aaa18151 return only eventId from findFirstOrLastOccurringEventId 2019-06-03 00:11:12 +02:00
Bruno Windels
7852f31f7e clear token on pagination when events start overlapping 2019-06-02 19:28:24 +02:00
Bruno Windels
0b637f656a timeline store requests should always be scoped to 1 fragmentId
as fragmentIds should not be sorted, they are a linked list
and that is what determines their order.
2019-06-02 19:27:40 +02:00
Bruno Windels
bb5f139355 fix fragmentId:0 being evaluated as falsy 2019-06-02 19:27:23 +02:00
Bruno Windels
4b5b90e199 fix another direction mismatch 2019-06-02 19:26:45 +02:00
Bruno Windels
c63d94947f fix persisting a gappy sync 2019-06-02 18:28:38 +02:00
Bruno Windels
ef5d2cfa08 chunk is in reverse-chronological order for backward pagination 2019-06-02 18:15:40 +02:00
Bruno Windels
a59014475b also swap logic of setter 2019-06-02 18:15:08 +02:00
Bruno Windels
bdad0ad86b fix some gap fill errors 2019-06-02 15:46:24 +02:00
Bruno Windels
d022608a1a it's entry.entryIndex, not eventIndex, but use helper method instead. 2019-06-02 15:24:27 +02:00
Bruno Windels
1ed3babfec fragment boundary is a gap if backwards(started) & previousToken 2019-06-02 15:01:14 +02:00
Bruno Windels
e339cb7321 more fixes, timeline is showing again 2019-06-02 14:59:30 +02:00
Bruno Windels
a1e527ccbc first round of fixes after running the app again in the browser! 2019-06-02 00:49:47 +02:00
Bruno Windels
3de86cdf33 obsolete comment 2019-06-01 18:32:32 +02:00
Bruno Windels
039bbe038c adjust tiles(collection) to entry changes 2019-06-01 18:29:02 +02:00
Bruno Windels
35d90a8535 ctor takes a bool, not Direction. But use helper fn instead. 2019-06-01 18:20:29 +02:00
Bruno Windels
2a128ed32c use SortedArray in Timeline, adjust loadAtTop to use TimelineReader 2019-06-01 17:39:23 +02:00
Bruno Windels
f8fbfbff9a implement reading n events from end of live fragment 2019-06-01 17:04:05 +02:00
Bruno Windels
3137f025c7 remove draft commented code 2019-06-01 16:44:58 +02:00
Bruno Windels
447b0aa03c allow adding fragments to comparer, instead of reloading from db.
This is a suboptimal implementation now,
but is the API we want to end up with. Readers and Writers
in persistence add fragments to the comparer when they become
aware of, create, or link up fragments.
2019-06-01 16:42:57 +02:00
Bruno Windels
fa4efe0132 rename Gap/SyncPersistence to Writer, in line with TimelineReader 2019-06-01 15:40:21 +02:00
Bruno Windels
784588440c WIP for fragment support 2019-05-19 20:49:46 +02:00
Bruno Windels
3324fd3afd split up persistFragmentFill method into smaller ones 2019-05-12 20:44:20 +02:00
Bruno Windels
10457611f9 whitespace 2019-05-12 20:26:46 +02:00
Bruno Windels
e3328f0fef add fragments store name 2019-05-12 20:26:32 +02:00
Bruno Windels
da5e8794ab lint 2019-05-12 20:26:20 +02:00
Bruno Windels
75100c1c60 adjust Timeline to changes, gap persister 2019-05-12 20:26:03 +02:00
Bruno Windels
a6b17cf25a first draft of persistFragmentFill 2019-05-12 20:25:41 +02:00
Bruno Windels
89bc0e1696 split up RoomPersister to SyncPersister
also rename stores to timelineEvents and timelineFragments
2019-05-12 20:24:06 +02:00
Bruno Windels
bf835ac01d create Entry classes and return fragment boundaries as entries as well
they can then be used for gap tiles.
2019-05-11 18:19:53 +02:00
Bruno Windels
2b510b24d9 adjust persister to fragments (untested) 2019-05-11 15:41:46 +02:00
Bruno Windels
152397a292 first impl of idb fragment store 2019-05-11 15:41:09 +02:00