[Mono-dev] XmlTextReader.cs patch (IBM bug workaround)

Eyal Alaluf eyala at mainsoft.com
Wed Apr 18 09:07:26 EDT 2007


Hi, Atsushi.

We have run extensive tests against IBM JIT and havn't seen other flows
of this kind. So we believe the case to be isolated despite the fact
that it looks like a more general scenario.
I definitely agree that we cannot have two many workarounds like this in
the code since such an approach is unmaintainable.
In this particular case I feel that we havn't bypass the limits of
maintainability and would like to accept this patch as a practical
solution to the problem we see.
Miguel, do you have any comments?

Eyal.

-----Original Message-----
From: mono-devel-list-bounces at lists.ximian.com
[mailto:mono-devel-list-bounces at lists.ximian.com] On Behalf Of Atsushi
Eno
Sent: 18 April 2007 14:54
To: Sasha Kogan
Cc: mono-devel-list at lists.ximian.com
Subject: Re: [Mono-dev] XmlTextReader.cs patch (IBM bug workaround)

I don't think we should accept such patches. Such a JIT bug will
happen to *every* classes in every libraries, and EVEN IF we
change the every code like this, issues will still happen on
any third-party libraries. Do you think you can ask everyone
to change his or her managed code just for this reason?

Atsushi Eno


Sasha Kogan wrote:
> 
> 
>  
> 
>  
> 
> ** /Sasha **
> 
> * From: * Sasha Kogan
> *Sent:* Monday, April 16, 2007 4:14 PM
> *To:* 'mono-devel-list at lists.ximian.com'
> *Subject:* XmlTextReader.cs patch (IBM bug workaround)
> 
>  
> 
> Please review the attached patch for XmlTextReader.cs. The patch is 
> actually a workaround for an IBM bug (seems to be a JIT compiler bug) 
> that causes java.lang.AbstractMethodError to be thrown upon access of 
> some virtual methods/properties within the class.
> 
> The workaround is to access such methods/properties indirectly, via
the 
> introduced 'This' property that just returns 'this' reference.
> 
>  
> 
> Alternativelly, we could wrap the patch in "#if TARGET_JVM", but since

> there're numerous places in the code where it should be applied, this 
> approach would dirty the code too much.
> 
>  
> 
> ** /Sasha **
> 
>  
> 
> 
>
------------------------------------------------------------------------
> 
> _______________________________________________
> Mono-devel-list mailing list
> Mono-devel-list at lists.ximian.com
> http://lists.ximian.com/mailman/listinfo/mono-devel-list

_______________________________________________
Mono-devel-list mailing list
Mono-devel-list at lists.ximian.com
http://lists.ximian.com/mailman/listinfo/mono-devel-list



More information about the Mono-devel-list mailing list