directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Emmanuel Lecharny <elecha...@gmail.com>
Subject Re: Disable Kerberos modules and tests (was: Build failed in Hudson: dir-studio-jdk15-ubuntu-deploy #97)
Date Sat, 20 Nov 2010 11:28:38 GMT
On 11/20/10 11:52 AM, Felix Knecht wrote:
> On 11/20/2010 11:31 AM, Emmanuel Lecharny wrote:
>> On 11/20/10 11:03 AM, Felix Knecht wrote:
>>>>> Thoughts?
>>>> I can try to spend some time this week-end on the failing tests, just
>>>> because it's annoying anyway.
>>>
>>> I already found some [1]. The problem for those was [2].
>>>
>>> Not sure, but maybe more failing tests are also caused by [2]. When
>>> just reverting [2] most of the tests are working again.
>>
>> It's most certainly an issue with the way we deal with BitString. Th pb
>> is that the new codec uses the modified version, so if I revert, it will
>> break the new codec.
>>
>> I'll try to understand why the old code is not happy with the perfect
>> code I injected in the kerberos part ;)
>
> My guess is that the getBytes(int xxx) has something to do with it. In 
> comparison to the code before it adds some additional information in 
> the [0] index (A first byte containing the number of unused bits is 
> added) whereas in the old code it just started without additional 
> information.
And this extra byte is mandatory, so I'm pretty sure that the old code 
does not handle it. I have *no* idea how possibly the kerberos server 
could have worked without this extra byte. Maybe the RFC says something 
about it, like 'the first clients weren't able to handle the BISTRING 
DER encoded as expected, so we accept the BITSTRING without the extra 
byte'. I have to check that...


-- 
Regards,
Cordialement,
Emmanuel L├ęcharny
www.iktek.com


Mime
View raw message