Closed Bug 343792 Opened 14 years ago Closed 13 years ago

Freeze (hang) after import of .ics file which has INTERVAL=0 in RRULE


(Calendar :: Internal Components, defect, critical)

Not set


(Not tracked)



(Reporter: ctuffli, Assigned: mvl)


(Keywords: dataloss, hang, Whiteboard: [libical-upstream?])


(3 files, 1 obsolete file)

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv: Gecko/20060508 Firefox/
Build Identifier: Mozilla Thunderbird version 3 alpha 1 (20060706) with Lightning 0.1+ (build 2006070507)

Exported an Outlook calendar to an .ics file with Outport ( and tried to import the resulting file. Lightning  started the import but became unresponsive (100% CPU utilization) before the import completed. Killing Thunderbird and restarting caused the application to again go into an unresponsive state and using nearly 100% of the CPU cycles. The only recovery is to delete the profile and restart Thunderbird.

Narrowed down  the problem to a single line:

Changing the INTERVAL value to 1 causes the import to suceed. The RFC specifies "positive integer" values, so it seems like Outport is outputing an incorrect RRULE, but it would be great if Lightning didn't hang.

Reproducible: Always

Steps to Reproduce:
1. Calendar > Import ...
2. Select the .ics file and click Open
Actual Results:  
Application hangs and the CPU utilization approaches 100%

Expected Results:  
Import the file and/or flag the error
This is a sanatized subset of the original file that causes Lightning to hang.
Ever confirmed: true
I have been having a similar issue, but in subscribing to a remote calendar. Unfortunately, my import problems went beyond the simple INTERVAL=0 issue, and I found that, in fact, semicolons were the problem, not the INTERVAL=0...

By removing all semicolons from the RRULE: line, the import of my test file succeded. Alternately, one of the failing lines from my test file imported successfully when I placed a semicolon at the end of the line (when other semi's were already present). Oddly, some RRULE's imported just fine with semicolons in the line, but without one on the end.

Also, I found that the single instance of INTERVAL=0 in my file imported correctly with the semicolons removed.

So, effectively, the RRULE parser just needs to be improved a bit to not hang on the presence or absence semicolons.

I'm running:
Thunderbird version 2 beta 1 (20061013)
Lightning 0.3
(In reply to comment #2)
Adding/removing semicolons changes the meaning of the RRULE in significant and unexpected ways.  Doing so is not considered a good solution to the problem.
bottom line is that Lightning shouldn't hang thunderbird in the event of a malformed file.
blocking 0.5?
Flags: blocking-calendar0.5?
Sunbird is also affected. Updating severity to critical because the only way to get rid of the problem is by removing entire storage.sdb file -> dataloss.
Severity: normal → critical
Component: Lightning Only → Internal Components
Keywords: dataloss, hang
OS: Windows XP → All
QA Contact: lightning → base
Summary: import of .ics file hangs Lightning if INTERVAL=0 → Freeze (hang) after import of .ics file which has INTERVAL=0 in RRULE
Version: unspecified → Trunk
Attached patch fool-proof libical (obsolete) — Splinter Review
This patch makes libical more foolproof, by ignoring interval=0.
Attachment #247727 - Flags: first-review?(daniel.boelzle)
Comment on attachment 247727 [details] [diff] [review]
fool-proof libical

>+CFLAGS += -g


leftovers from your debugging session?
(If not, I don't think we should compile with debug or abort the app on ical errors.)

>-	    parser.rt.interval = (short)atoi(value);
>+	    short v = atoi(value);
missing cast from int to short here, avoiding a compiler warning.

>+	    if (v) {
>+		parser.rt.interval = (short)atoi(value);
>+	    }
better just
		parser.rt.interval = v;

>-    icaltimezone_parse_zone_tab ();
>+//    icaltimezone_parse_zone_tab ();
Why this?
Attached patch patch v2Splinter Review
Patch now without the leftover stuff in other files. Sorry for the mess.
Also made the check a little bit more strict: also ignore negative values. Just to be sure.
Attachment #247727 - Attachment is obsolete: true
Attachment #248420 - Flags: first-review?(daniel.boelzle)
Attachment #247727 - Flags: first-review?(daniel.boelzle)
Attachment #248420 - Flags: first-review?(daniel.boelzle) → first-review+
Assignee: nobody → mvl
patch checked in
Closed: 13 years ago
Resolution: --- → FIXED
No freeze during import using Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv: Gecko/20061215 Thunderbird/2.0b1 and Lightning/0.4a1 (2006121603)

No freeze during import using Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9a1) Gecko/20061217 Calendar/0.4a1

Flags: blocking-calendar0.5?
If possible there should be a regression test case in Litmus for this issue.
Flags: in-litmus?
imo, there should be a unit-test instead. This can perfectly be tested automatically. No need for manual litmus labour.
Flags: in-litmus? → in-testsuite?
Attachment #374584 - Flags: review?(philipp)
Comment on attachment 374584 [details] [diff] [review]
xpcshell testcase

Looks good, r=philipp
Attachment #374584 - Flags: review?(philipp) → review+
Flags: in-testsuite? → in-testsuite+
Seems this is not in libical 0.46 yet, we should make sure this goes upstream
Whiteboard: [libical-upstream?]
You need to log in before you can comment on or make changes to this bug.