Reviewer needed: backport of "6792400: Avoid loading of Normalizer resources, for simple uses" into IcedTea6 HEAD

[Pavel Tisnovsky]

Jiri Vanek wrote:
> On 03/26/2012 04:02 PM, Pavel Tisnovsky wrote:
>> Hi,
>>
>> I'd like to push following backport to IcedTea6 HEAD (and to OpenJDK6
>> too - will do parallel request):
>> "6792400: Avoid loading of Normalizer resources, for simple uses".
>>
>> This backport fixes following IT6 bug:
>> http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=882
>>
>> hg diff created against recent IcedTea6 HEAD is stored in an attachment.
>>
>> ChangeLog entry:
>> 2012-03-26  Pavel Tisnovsky<ptisnovs at redhat.com>
>>
>>          *
>> patches/openjdk/6792400-Avoid_loading_Normalizer_resources.patch:
>>          Backport of: 6792400: Avoid loading of Normalizer resources
>>          for simple uses
>>          * NEWS: Mention backport.
>>          * Makefile.am: Updated
>>
>> This backport works correctly - passed all regular tests.
>>
>> Can anybody please review this change?
>>
>> Thank you in advance,
>> Pavel
> 
> Hi!
> 
> I have walked across the change set and I have few questins:

Thanks Jiri!

> 
> -Sun bugzilla is mentioning, that it is "just" performance improvement.
>  ?
> -As far as I know, the Jira was falling _inside_ normalize method(s), so
> this fix looks like not fixing the issue, but just to walk-around for
> just-ascii texts.

The method for string normalization is called from the static block and it throws
exception when called with ASCII-only string. There's nothing which should be
normalized in such strings ;-)

>  Am I right?
>  Was the jira failures caused just-by-ascii input files?

Yes. To be precise - it does not work with test files but only with text strings.

> -I know that you have dig around to find "some fixing backport", so
> there is none fixing just corrupted normalizer?

It's not needed at all to fix bug #882 and as you know nobody were happy with
1MB patch ;-) I could try to find a fix for the normalizer, but it won't be
immediate which will slows us down.

> -Also I know that the reproducer was quite complicated, but still, can
> you ssy little bit more about it?

Hmm I could try to export a ST from the test reports.

> 
> I do not believe there is direct fix for this issue, and unless there is
> one, and this is the last of backport for jira issue this should go in 
> no metter how much I dont like it :)

Heh what's wrong with this patch. After all it'll make IcedTea6 quite quicker :-)

> 
> Thanx for dooing it.
>   J.