harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Pavel Rebriy" <pavel.reb...@gmail.com>
Subject Re: [drlvm][verifier] verifier behavior is not compatible with RI
Date Wed, 31 Jan 2007 13:49:16 GMT
The patch is applied to HARMONY-2905 [1] and committed to svn r501839 by
Gregory Shimansky.

[1] https://issues.apache.org/jira/browse/HARMONY-2905

On 30/01/07, Pavel Rebriy <pavel.rebriy@gmail.com> wrote:
>
> Let I try to explain how verifications are working. There are 3 modes of
> DRLVM verifier:
>
> 1. Default mode – without any VM control options
>
> In this mode DRLVM verifier should be compatible with RI default mode.
>
> 2. With options "–verify" or "–Xverify"
>
> This mode is supposed to verify more strict checks of class file format
> (such as correct naming of fields and methods in constant pool), besides
> expects verification of classes loaded by bootstrap classloader.
>
> 3. With options "-Xverify:all"
>
> This mode is supposed to more strict verification of classes (such as
> interface classes' verification).
>
> Because reference VM doesn't support verification of 'uninitialized this'
> class reference under exception handler protection I'm going to move it to
> mode 3, so check this verification only if option –Xverify:all is set.
>
> Such behavior should be compatible with reference VM.
>
>
> On 22/01/07, Geir Magnusson Jr. <geir@pobox.com> wrote:
> >
> >
> > On Jan 19, 2007, at 12:51 PM, Gregory Shimansky wrote:
> >
> > > Ivan Popov wrote:
> > >> I'd like to discuss the problem with Eclipse TPTP profiler working
> > >> with DRLVM, which is described in HARMONY-2905 [1].  The problem is
> > >
> > > There is always a workaround to verifier exceptions. You can run
> > > the program with -Xverify:none to disable verifier completely.
> > > Turning this particular check is simple too. The question is
> > > whether this should be a default mode in VM or whether it should be
> > > enabled by some special option which doesn't disable all other
> > > verifications.
> > >
> >
> > I'm also darn curious why Sun's verifier doesn't have the problem...
> >
> > geir
> >
> >
>
>
> --
> Best regards,
> Pavel Rebriy
>



-- 
Best regards,
Pavel Rebriy

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message