ZP-1157 / similar errors with Caldav/Carddav/IMAP and Windows 10 mail/people/calendar

cpitchford

I’ve had a look at this commit:

https://stash.z-hub.io/projects/ZP/repos/z-push/commits/14c1b661aff7cbeabe0d885db186c7e2b76e0ccd

If I back it out, my test user is able to sync, though my regular user still hits the same unexpected synckey value.

It definitely changes, though when this commit is removed. I’m wondering if it introduced another problem, or it’s not entirely fixed the bug it was addressing?

I’ll play more tomorrow!

Sebastian

I looked at your log but I can’t figure the unexpected synckey value from it. I tried WP but I can’t make it happen.
The only think that seems sure is that

 self::$deviceManager->GetStateManager()->SetSyncState($spa->GetNewSyncKey(), $state, $spa->GetFolderId());

is being called in lib/request/sync.php line 1248 (in develop).
This is a bit odd because $spa->HasNewSyncKey() from a few lines up shouldn’t really be true at this point.
It could be that this is related to SyncParameters->uuidNewCounter which is checked there.
You could debug this a little bit by added a few more debug outputs around these lines and printing values.
It could also be that there is another UID in the game which was checked before, because in your log in the next request another state is missing:

SyncParameters->GetLatestSyncKey(): '{54324d90-32e6-4816-ae74-efed41ecb7ba}1'
StateNotFoundException: FileStateMachine->GetState(): Could not locate state '/var/lib/z-push/1/e/3d8750c317150576af476d9dbd631ee1-54324d90-32e6-4816-ae74-efed41ecb7ba-1'

It should read counter 2 here, because 1 was already deleted. This would indicate that the -fd file is not correctly updated reflecting to an old state.

Sebastian

Don’t worry too much about the user agent. WP doesn’t send it in all requests and there we just print the header (which is not available).
It’s because that’s an unexpected exception that is not being catched.

Manfred

Hi cpitchford,

@cpitchford said in ZP-1157 / similar errors with Caldav/Carddav/IMAP and Windows 10 mail/people/calendar:

Ah that’s interesting!!

I think I’ve spotted the problem, will test and report back… but:

My log file was actually filtered (sensitive terms and names changed to protect the innocent!)

Lets pretend my calendar is actually called “Testing” not “testing”

So, I’d already defined CALDAV_PERSONAL in my config:
define('CALDAV_PERSONAL', 'Testing');
However, there are two (identical) if statements that check the calendar:
if (defined('CALDAV_PERSONAL') && strtolower(substr($id, 1)) == CALDAV_PERSONAL) {
The if statement won’t match as its only converting the caldav calendar name, not the defined constant

So, I propose (as a patch) changing the if statements in caldav.php into:
if (defined('CALDAV_PERSONAL') && strcasecmp(substr($id, 1), CALDAV_PERSONAL) == 0) {
So that the names are compared case-insensatively

(the alternative is, perhaps, a comment in the caldav config file explaining CALDAV_PERSONAL must be entirely lower case.

The caldav and cardav backends were contributed by the community, so feel free to open a JIRA ticket, commit the change and send us a pull request.
We have a wiki page on how to contribute to Z-Push: https://wiki.z-hub.io/display/ZP/Development+guidelines

Manfred

cpitchford

@Manfred

Hi Manfred,

I’m hoping I’ve done things correctly and created ZP-1244 for the case-matching problem. It’s looking good on my set up so fingers crossed!

I’m going to crack on with inserting more debug lines into the code to see if I can watch what’s happening to this missing states

Sebastian

Patch looks good to me. I’ve set the target version to 2.3.8.

Sebastian

Your patch is merged into develop and is also in 2.3.8beta0+22 (just published).

Thanks!

cpitchford

I’ve narrowed down the original problem (now running 2.3.8beta0)

I’ve set up a test account:

zero emails
zero calendar entries
zero tasks
one contact

When I set up the account in MS Mail, it will sync, once. If I press “refresh” it faults (and Z-Push returns 500)

I can’t figure out why [StateManager]->newStateCounter is either unset or “”

It should be set to either 1 or 1+the current counter if [StateManager]->GetNewSyncKey is called…

But for this StateManager, it appears GetNewSyncKey was not called.

So if GetNewSyncKey was not called, why would it need to call [StateManager]->SetSyncState??

I’ve an updated log of configuring the client, syncing once, retrying and hitting the error

https://maildebug.intrepid.co.uk/capture3.log

I’m wondering if it is related to loop detection? Or possibly a php issue.

Sebastian

Hey Chris,

thanks for debugging this, your log gave some good indications what was going on. Disabling the FolderStat support I could also reproduce this with the Kopano backend.

I’ve created a ticket and was able to fix the issue. The states were being confirmed which leads at a certain point to the deletion of the old state, but the folderdata (SyncParameters) are not saved/updated correctly.

https://jira.z-hub.io/browse/ZP-1247

The fix is already merged into develop. Could you test that?

Cheers,
Sebastian

Sebastian

There is also ZP-1248 which adds the windows mail app to the “long timeout device list”. It’s also in develop and will require to update your main config.

cpitchford

It’s looking great!!

There were no problems syncing my test account (with nothing but one contact) and it’s now synced my regular personal email!

Would you like me to capture the logs as before, for comparison?

I’m going to go study the diff to see what you’ve changed, I’m keen to know more about how this works unde the hood!

Sebastian

No, no need for other logs at this point.
Let me know if you need anything else.

Cheers,
Sebastian