Re: Isolating a bug?

FROM : Gregory Weston
DATE : Wed Apr 06 15:54:39 2005

On Wednesday, April 06, 2005, at 07:41AM, Ondra Cada <<email_removed>> wrote:

>On 6.4.2005, at 12:46, Gregory Weston wrote:
>

>> Exception:  EXC_BAD_ACCESS (0x0001)
>> Codes:      KERN_PROTECTION_FAILURE (0x0002) at 0x00000006
>>
>> Thread 0 Crashed:
>> 0  com.apple.CoreFoundation     0x901939cc CFHash + 0x18
>> 1  com.apple.CoreFoundation     0x90192d28 __CFDictionaryFindBuckets1b +
>> 0x70
>> 2  com.apple.CoreFoundation     0x90192184 CFDictionaryGetValue + 0xec
>> 3  com.apple.CoreFoundation     0x901b0fb0
>> _CFApplicationPreferencesCreateValueForKey + 0x6c
>> 4  com.apple.Foundation         0x909fb38c -[NSUserDefaults
>> objectForKey:] + 0x98
>> 5  my.program.signature         0x00008658 -[MyApplication
>> methodTakingAString:] + 0x5c
>>
>> The routine identified in line 5 is called repeatedly (although not
>> rapidly) while the program is running - it had been invoked hundreds
>> of times without error during this session. The argument to
>> objectForKey: at line 4 is freshly-created from
>> NSString>>stringWithFormat:. I'm not sure where to start looking for
>> the cause of a problem that manifests five invocations into external
>> code. Advice?

>
>For starters, log the "freshly-created from NSString>>stringWithFormat"
>just before the objectForKey: gets called. There is still a chance it
>gets something messed up somehow...


In the abstract that's reasonable, but:
1) The string is created on the line before the objectForKey: message is sent. There's not a whole lot of opportunity for corruption. To answer someone else's comment, I don't release it and I don't cache it such that I might be releasing it unintentionally later.
2) As noted, this has been deployed for about a year. As not noted, but worthy of note, it runs on a _lot_ of machines and this routine may run thousands of times a day. That's a lot of logging for something that may not show up for another year.

>If that does not help, I kind of fear the problem *may* be elsewhere,
>in an uninitialized pointer or something alike, whose result are
>mangled in-memory data.


My fear, too.