369885 - Can't navigate, View fails on repeating events with modified occurrences

Status: RESOLVED FIXED

(Calendar :: Provider: ICS/WebDAV, defect)

Description

[matthaeus123]

With the nightly branch builds of Calendar 0.4a1, the user is unable to navigate through months. The names of the months change but the dates do not change.

Reproduce:
1. Go to month view
2. try to go to next month.

What should have happend:
Calendar should have changed to next months events, and the events should be visible.

Error console: 2 Different errors

1.
Error: [Exception... "Component returned failure code: 0x80004005 (NS_ERROR_FAILURE) [calIItemBase.calendar]"  nsresult: "0x80004005 (NS_ERROR_FAILURE)"  location: "JS frame :: chrome://calendar/content/calendar-month-view.xml :: relayout :: line 426"  data: no]
Source File: chrome://calendar/content/calendar-month-view.xml
Line: 426

2.
Error: itd.item.calendar.uri has no properties
Source File: chrome://calendar/content/calendar-month-view.xml
Line: 426


Build ID:
Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.8.1.2pre) Gecko/20070209 Calendar/0.4a1

Comment 1

[matthaeus123]

Figured out this was caused by corrupted local calendars

Comment 2

[Stefan Sitter [:ssitter]]

I can reproduce the error from above using Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.8.1.2pre) Gecko/20070208 Calendar/0.4a1:

Steps to Reproduce:
1. Start Sunbird with clean profile
2. Create repeating event in ics calendar on local disc
3. Modify one occurrence of this event e.g. by changing start time and title
4. Restart Sunbird
5. Switch views

After startup I see the following error and the modified occurrence has no calendar color set:

Error: uncaught exception: [Exception... "Component returned failure code: 0x80004005 (NS_ERROR_FAILURE) [calIItemBase.calendar]"  nsresult: "0x80004005 (NS_ERROR_FAILURE)"  location: "JS frame :: chrome://calendar/content/calendar-view-core.xml :: setCSSClasses :: line 140"  data: no]

Switching to Month view produces the following error and I can't navigate as observed by matthaeus123:

Error: uncaught exception: [Exception... "Component returned failure code: 0x80004005 (NS_ERROR_FAILURE) [calIItemBase.calendar]"  nsresult: "0x80004005 (NS_ERROR_FAILURE)"  location: "JS frame :: chrome://calendar/content/calendar-month-view.xml :: relayout :: line 426"  data: no]

Opening the .ics file in another profile causes:

Error: item.calendar.uri has no properties
Source File: chrome://calendar/content/calendar-view-core.xml Line: 140

Import into storage calendar results in:

Error: itd.item.calendar.uri has no properties
Source File: chrome://calendar/content/calendar-month-view.xml Line: 426

This might be a regression from Bug 364841 and Bug 368490, further testing is required to find a regression range.

Comment 3

[Stefan Sitter [:ssitter]]

Attachment: ics testcase

The .ics file I created with Sunbird that causes the problems.

Comment 4

[matthaeus123]

I not get this message in the error-console


Error: itd.item.calendar.uri has no properties
Source File: chrome://calendar/content/calendar-month-view.xml
Line: 430

Comment 5

[matthaeus123]

(In reply to comment #4)
> I now getting this message in the error-console
> 
> 
> Error: itd.item.calendar.uri has no properties
> Source File: chrome://calendar/content/calendar-month-view.xml
> Line: 430
> 

Comment 6

[Stefan Sitter [:ssitter]]

Regression range testcase file from above with ics provider:

Works in 0.4a1 (2007011008) ok
Fails in 0.4a1 (2007011010) modified occurrences not displayed (bug 368490)
...
Fails in 0.4a1 (2007012803) modified occurrences not displayed (bug 368490)
Fails in 0.4a1 (2007012912) error in views (this bug)

So this indead looks like a regression from Bug 364841 that was covered by Bug 368490 the first time.

Comment 7

[Martin Schröder [:mschroeder]]

matthaeus123, please don't grant blocking-calendar0.5 (+) yourself, only ask for blocking-calendar0.5 (?). Granting will be done by the driver of the release.

Comment 8

[matthaeus123]

Oh, I'm sorry.

Comment 9

[Frédéric Buclin]

(In reply to comment #7)
> matthaeus123, please don't grant blocking-calendar0.5 (+) yourself, only ask
> for blocking-calendar0.5 (?). Granting will be done by the driver of the
> release.

You can prevent that by setting the grant group for this flag type. Ask someone with editcomponents and editgroups privs on b.m.o. ;)

Comment 11

[matthaeus123]

FYI this bug occurs on the trunk and nightly builds

Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9a3pre) Gecko/20070227


Also, I think that this should be set at a severity higher than normal. This is a pretty big bug.

Comment 12

[Stefan Sitter [:ssitter]]

I think the following code in calIcsParser.js is the cause for the error:

http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/calendar/base/src/calIcsParser.js&rev=1.2&mark=133#116

    var rid = item.recurrenceId;
    if (!rid) {
        unexpandedItems.push(item);
        if (item.recurrenceInfo) {
            uid2parent[item.id] = item;
        }
    } else {
        item.calendar = this;
        // force no recurrence info:
        item.recurrenceInfo = null;
        excItems.push(item);
    }

Previously, in the old calICSCalendar.js implementation "this" in 
"item.calendar = this;" refered to the calICSCalendar object derived from calICalendar.

But in the new implementation "this" refers to the calIcsParser object that doesn't implement calICalendar. Therefore the later call to access one of the calICalendar properties fails.

Comment 13

[Stefan Sitter [:ssitter]]

Testing showed that removeing the line in calIcsParser that sets .calendar and adding a loop in calICSCalendar that sets .calendar for all items afterwards doesn't work.

The reason is that the .calendar property was set for the exception items only. Later the exception item was stored in recurrenceInfo property of the parent item.

http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/calendar/base/src/calIcsParser.js&rev=1.2&mark=133,145-149#130

The calIcsParser returns only the list of parent items. For this items the .calendar property is set during adding to the memory calendar. For the exception items the .calendar property is not set - hence the error.

Comment 14

[Stefan Sitter [:ssitter]]

Attachment: rev0 - fix for calICSCalendar

This fixes calICSCalendar only by setting the .calendar property for all items and exceptions after retrieving the items from the calIcsParser.

Note: A similar solution will be required for all other users of calIcsParser, e.g. calIcsImporter.

Comment 15

[Michiel van Leeuwen (email: mvl+moz@)]

Comment on attachment 257226 [details] [diff] [review]
rev0 - fix for calICSCalendar

>Index: mozilla/calendar/providers/ics/calICSCalendar.js
>+                            exception.calendar = this;
>                 this.mMemoryCalendar.adoptItem(item, null);

When you call adoptItem with an item, the caller should not have to worry about setting .calendar on the item. The provider is supposed to be doing that.
So your loop should be moved into the memory calendar, in the adoptItem method.

Comment 16

[Stefan Sitter [:ssitter]]

Attachment: rev1 - fix for calICSCalendar [checked in]

Move the loop into calMemoryCalendar::adoptItem() as requested

Comment 17

[Michiel van Leeuwen (email: mvl+moz@)]

Comment on attachment 257268 [details] [diff] [review]
rev1 - fix for calICSCalendar [checked in]

r=mvl

Comment 18

[Stefan Sitter [:ssitter]]

This patch from Comment #16 only fixes the IcsProvider and the import into IcsCalendar.

But after import into storage calendar the problem still exists until Sunbird/Thunderbird is restarted. I could not test how other providers behave.

Leave the bug open after checkin or file a followup bug?

Comment 19

[Matthew (lilmatt) Willis]

Leave open.

Comment 20

[Matthew (lilmatt) Willis]

Patch checked in on MOZILLA_1_8_BRANCH and trunk.

Comment 21

[matthaeus123]

The patch seems like it didn't fix the problem. And has caused a fairly critical regression. I am getting this error now.

Error: aDate is not defined
Source File: chrome://calendar/content/calendar-decorated-base.xml
Line: 104



Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9a3pre) Gecko/20070308 Calendar/0.6a1

Comment 22

[Stefan Sitter [:ssitter]]

(In reply to comment #21)
No. This is a regression from Bug 370836 and already fixed. See Bug 373187 Comment #2.

Comment 23

[matthaeus123]

ok

Comment 24

[Stefan Sitter [:ssitter]]

qawanted: After the ics calendar being fixed: What's the current failure mode when importing testcase to e.g. caldav, wcap, gdata calendar?

Comment 25

[Bruno Browning]

(In reply to comment #24)
> qawanted: After the ics calendar being fixed: What's the current failure mode
> when importing testcase to e.g. caldav, wcap, gdata calendar?
> 

I imported the testcase to a caldav calendar using current Linux/branch nightly and was able to navigate (which is to be expected since the caldav provider does not yet use calIcsParser.js). No js console errors.

Comment 26

[Andreas Treumann]

I checked this issue in the latest lightning-wcap build and it works for me, too.

Comment 27

[Matthew (lilmatt) Willis]

Given that this apparently doesn't affect usability of WCAP and CalDAV, I'm removing the blocking flag from it.

Comment 28

[Matthew (lilmatt) Willis]

Not going to make the 0.5 train.

Comment 29

[Stefan Sitter [:ssitter]]

Resolving this bug as FIXED based on the previous comments.

Please file new bug reports that references this bug if there appear issues with other providers too.