Request for review: 7084245: Update usages of InternalError to use exception chaining

Sebastian Sickelmann sebastian.sickelmann at gmx.de
Mon Aug 29 07:58:53 UTC 2011


Am 29.08.2011 02:02, schrieb David Holmes:
> On 29/08/2011 5:35 AM, Sebastian Sickelmann wrote:
>> Hi, here is a webrev[1] for some cleanup that i want to integrated in
>> tl-repositories.
>>
>> Alan Bateman had scanned the changes and gave me some good input[3] for
>> further discussion here:
>>
>> The changes to java.util.concurrent should go through Doug Lea's 
>> upstream
>> CVS. Alan told me that Chris Hegarty is on this topic already. The 
>> suggested
>> changes for this is here[2].
>
> These changes are somewhat pointless. We don't need to set the cause 
> because it will never happen. The classes are cloneable so 
> CloneNotSupportedException can not be thrown. If it is then something 
> really bizarre has happened hence we throw InternalError. The site of 
> the InternalError tells us why it was thrown.
>
> David
> -----
Thanks for your input. I have myself thought many times about the 
effects of these changes. It changes almost nothing, especially because 
it chains to InternalError which should not be thrown in the most cases. 
So chaining almost costs us zero runtime-costs but in really bizarre 
situations it gives slightly more information what happend.
This cleanup was the first step in a series of cleanup work i want to do 
in topic of exception-chaining. I chose it(InternalError) as the first 
step because it it has so less side-effect(longer chains, runtime 
performance while serializing and printing, later GC of causes,...) 
because it should mostly never happen. So i had the change to learn 
using the tools(mercurial, openjdk build files, webrev, contribution 
process).
The only problem i actually see with these small-changes (where nothing 
really bad/wrong is cleanuped) is the workload i generate for the 
reviewers, supporters, mailinglist-readers and if its worth doing it anyway.

So let me know if the complete cleanup of exception chaining should be 
considered to stop. It's ok if it's so. I have learned enought in using 
the tools and i can start searching for more interessting things to do. 
I only keep going because i said i will do it. It the point in time (Aug 
8) i guessed that there is much to do if I start with it. But i haven't 
imagine that's so much assiduity work.

-- Sebastian



More information about the core-libs-dev mailing list