On Apr 26, 2010, at 8:15 PM, David Smith wrote:
> On Apr 26, 2010, at 5:47 PM, Evan Schoenberg, M.D. wrote:
>
>> So LMX is supposed to be able to take data in chunks, from the end of the
>> file. It turns out that this doesn't work if a chunk breaks up an entity,
>> though.
>>
>> A chat transcript, submitted by lhirlimann to me via email by request,
>> exhibits this behavior on my machine, when read by
>> -[DCMessageContextDisplayPlugin contextForChat:], since that method reads in
>> 4*getpagesize() byte chunks and that number of bytes from the end is the
>> middle of an entity.
>>
>> Specifically, the resulting chunk #1 starts with 'apos'. The resulting
>> chunk #2 ends with '&'.
>>
>> The result is an assertion failure:
>> *** Assertion failure in -[LMXParser parseEntityReference:],
>> /Volumes/Huffman/Projects/lmx-sholt/LMXParser.m:605
>> "entity name was empty! currentIndex is 5193 (&) out of 21579 bytes"
>>
>> It's not clear to me how to proceed with a fix. Do you (yes, you!) have any
>> ideas?
>>
>> I don't want to post the user's chat log to a public mailing list; if you're
>> an existing developer and would like it in order to easily see the behavior,
>> email me off-list.
>>
>> This took way too long to hunt down, though it's a fairly obvious problem in
>> retrospect.
>>
>> Cheers,
>> Evan
>
> Silly hackish fix: @try { thing that throws assertion failure } @catch (...)
> { read a few more bytes so we're not in mid-entity and try again }
Better than my @try and (@)fail, though also more complex. Can you implement
it? :)
-Evan
