harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Asaf Yaffe <asaf_ya...@yahoo.com>
Subject Re: [drlvm][verifier-ext] launching a profiler
Date Sun, 11 May 2008 07:13:41 GMT
Hi Alexei,

To turn-on logging, please define the following environment variables (example is using Windows
syntax):

set MARTINI_LOGGER_DIRECTORY=<full path to an existing directory>
set MARTINI_LOGGER_LOG_LEVEL=5

Then start the profiler. A log file named MLog*.log will be generated in MARTINI_LOGGER_DIRECTORY.
Please post it here.

P.S: What O/S are you using for testing?

Thanks,
Asaf



----- Original Message ----
From: Alexei Fedotov <alexei.fedotov@gmail.com>
To: dev@harmony.apache.org
Sent: Friday, May 9, 2008 1:21:00 PM
Subject: Re: [drlvm][verifier-ext] launching a profiler

Vasily, Asaf, all,

We have discussed incorrect stack maps generated for a constructor
exception handler when the range invokes another constructor of the
same class. The Java 6 verifier specification (JSR 202) has two
paragraphs describing this case ending with a claim for "not this
case". I started using your profiler to understand what RI expected in
the stack map for this case (uninitialized this? or top?). I have
instrumented DRLVM to generate a required stack map and now I am
trying to launch your profiler.

I have composed helpful pieces of advice in one place [1], followed
the path, and got the following problem:
        agent library failed to init: JPIBootLoader.
Do you have any experience why this may happen? Is there any way to
turn on debugging in your agent?

Thanks!

[1] http://wiki.apache.org/harmony/Verifier_Extension

On Sun, Apr 20, 2008 at 8:33 PM, Vasily Levchenko
<vasily.v.levchenko@gmail.com> wrote:
> I've mentioned patches attached to
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=148629
>
> On Sun, Apr 20, 2008 at 8:30 PM, Vasily Levchenko <
> vasily.v.levchenko@gmail.com> wrote:
>
>> Hello Alexei,
>> Currently java 6 support isn't integrated in Probekit and BCI engine
>> because its current implementation rely on Harmony verifier are going to
>> release ;). Patches to BCI introduces implementation of java 6 support and
>> code for initialization of dynamic agent using instrumentation like Call
>> Graph, Heap, thread and Dynamic Probekit profiler agent. Static
>> instrumentator wasn't supposed to be introduced in this work. But I suppose
>> it's the simplest way to verify work of extension of verifier on
>> instrumented code.
>>
>>
>> On Sun, Apr 20, 2008 at 11:01 AM, Alexei Fedotov <alexei.fedotov@gmail.com>
>> wrote:
>>
>> > Hello Vasily,
>> >
>> > I'm trying to understand how to made the simple verifier extension
>> > test from your static instrumentator. Or, in other words, I'm trying
>> > to understand the verifier extension interface and how to reproduce
>> > your bug [1]. I downloaded the probekit sources from dev.eclipse.org
>> > and tried to understand how do they use a verifier extension:
>> >
>> >  $  cvs -d :pserver:anonymous@dev.eclipse.org:/cvsroot/tptp co
>> > platform/org.eclipse.hyades.probekit
>> >  $ grep -rI recompute_stackmaptable platform/
>> >
>> > I cannot find anything. I started to think that your patch to the bug
>> > report [2] may contain stack map re-computation example and checked
>> > your attachments. But it seems that the recompute_stackmaptable
>> > function is missed there as well. Could you please give more
>> > directions? For example, could you point to the part of your work
>> > related to the stack map re-computation? I also wonder if
>> > recompute_stackmaptable is a part of public interfaces, or I should
>> > look for something else.
>> >
>> > Thank you in advance,
>> > Alexei
>> >
>> > [1] https://issues.apache.org/jira/browse/HARMONY-5764
>> > [2] https://bugs.eclipse.org/bugs/show_bug.cgi?id=148629
>> >
>> > On Fri, Apr 18, 2008 at 7:31 PM, Vasily Levchenko
>> > <vasily.v.levchenko@gmail.com> wrote:
>> > > Hello folks,
>> > >  Have you got any updates about commitment of
>> > >  H-5750<https://issues.apache.org/jira/browse/HARMONY-5750>.
>> > >
>> > >
>> > >  About testing.
>> > >  We've discussed it with some folks, but I don't know how it complex
>> > for
>> > >  testing system of Harmony.
>> > >  Actually the functionality we need is used for recalculating stack
>> > maps
>> > >  after instrumentation. There is a subproject of TPTP called Probekit
>> > that
>> > >  injects probes into compiled code. But for re-calculation requires
>> > valid JNI
>> > >  pointer (you can find some details in
>> > >  https://bugs.eclipse.org/bugs/show_bug.cgi?id=148629). So minimize
>> > test it
>> > >  possible re-use static instrumentator with introduced in the same
>> > bugzilla
>> > >  java6 support but for static instrumentation.
>> > >
>> > >  Is it ok with you?
>> > >
>> > >  On Fri, Apr 18, 2008 at 10:34 AM, Alexei Fedotov <
>> > alexei.fedotov@gmail.com>
>> > >  wrote:
>> > >
>> > >
>> > >
>> > >  > Nathan, thanks for a question!
>> > >  >
>> > >  > > Is VS6 needed to appropriately test this issue [1]? [...] are
>> > there
>> > >  > specific tests that could be run to get a general
>> > >  > > assurance of the passivity?
>> > >  >
>> > >  > I was asking Vasily to prepare at least one stand-alone test to be
>> > >  > included (by me) in a Harmony test base. While the whole TPTP
>> > requires
>> > >  > VC6, I believe I will be able to recompile the only test with a
>> > newer
>> > >  > compiler.
>> > >  >
>> > >  > With best regards, Alexei
>> > >  >
>> > >  >
>> > >  >
>> > >  > On Fri, Apr 18, 2008 at 5:27 AM, Nathan Beyer <ndbeyer@apache.org>
>> > wrote:
>> > >  > > Is VS6 needed to appropriately test this issue [1]? I'm not
>> > intimate
>> > >  > with
>> > >  > >  the verifier; are there specific tests that could be run to
get
>> > a
>> > >  > general
>> > >  > >  assurance of the passivity?
>> > >  > >
>> > >  > >  -Nathan
>> > >  > >
>> > >  > >  [1] https://issues.apache.org/jira/browse/HARMONY-5750
>> > >  > >
>> > >  > >  On Thu, Apr 17, 2008 at 2:22 PM, Vasily Levchenko <
>> > >  > >
>> > >  > > vasily.v.levchenko@gmail.com> wrote:
>> > >  > >
>> > >  > >
>> > >  > > > Greetings,
>> > >  > >  > We've finally established source base and launched our
test,
>> > >  > demonstrating
>> > >  > >  > stability and reliable of verifier code. would you mind
to
>> > initiate
>> > >  > with
>> > >  > >  > releasing milestone HDK-M5.5_Eclipse_TPTP?
>> > >  > >  >
>> > >  > >  > As I've mentioned earlier but wasn't able point to JIRA
(
>> > >  > >  > https://issues.apache.org/jira/browse/HARMONY-5750) we're
>> > extremely
>> > >  > >  > interested this patch to be included.
>> > >  > >  >
>> > >  > >  > On Mon, Apr 14, 2008 at 9:15 AM, Stepan Mishura <
>> > >  > stepan.mishura@gmail.com<
>> > >  >
>> > https://mail.google.com/mail?view=cm&tf=0&to=stepan.mishura@gmail.com>
>> > >  > >  > >
>> > >  > >  > wrote:
>> > >  > >  >
>> > >  > >  > > On 4/14/08, Vasily Levchenko <vasily.v.levchenko@gmail.com<
>> > >  >
>> > https://mail.google.com/mail?view=cm&tf=0&to=vasily.v.levchenko@gmail.com
>> > >  > >>
>> > >  > >
>> > >  > > > wrote:
>> > >  > >  > > > On Mon, Apr 14, 2008 at 8:32 AM, Stepan Mishura
<
>> > >  > >  > > stepan.mishura@gmail.com<
>> > >  >
>> > https://mail.google.com/mail?view=cm&tf=0&to=stepan.mishura@gmail.com>
>> > >  > >
>> > >  > >
>> > >  > > > >
>> > >  > >  > > > wrote:
>> > >  > >  > > >
>> > >  > >  > > > > Hi folks,
>> > >  > >  > > > >
>> > >  > >  > > > > As I understood the thread it is doable
to make interim
>> > release
>> > >  > >  > > > > targeted to assist inclusion of Harmony
verifier to the
>> > nearest
>> > >  > >  > > > > Eclipse TPTP release. Let me share my understanding
of
>> > the
>> > >  > request
>> > >  > >  > and
>> > >  > >  > > > > a possible way to resolve it (please correct
me if I'm
>> > wrong):
>> > >  > The
>> > >  > >  > > > > Eclipse team needs an 'official' (i.e.
published on the
>> > >  > web-site as
>> > >  > >  > > > > milestone build) Harmony release. The Eclipse
team only
>> > >  > interesting
>> > >  > >  > in
>> > >  > >  > > > > changes in verifier since M5 so the main
criteria for
>> > the
>> > >  > interim
>> > >  > >  > > > > release is no regressions in verifier functionality
>> > (i.e. I
>> > >  > assume
>> > >  > >  > > > > that not critical regressions are acceptable
for interim
>> > >  > release. I
>> > >  > >  > > > > believe that is important for having a
shorten
>> > >  > freeze/test/release
>> > >  > >  > > > > period for the interim release)
>> > >  > >  > > > >
>> > >  > >  > > > > So I think we may consider:
>> > >  > >  > > > > - making sure that all artifacts required
are in place
>> > (i.e.
>> > >  > >  > committed
>> > >  > >  > > > > to the trunk)
>> > >  > >  > > > > - declaring short code freeze
>> > >  > >  > > > > - running testing cycle to see if there
are any issues
>> > with
>> > >  > verifier
>> > >  > >  > > > > and overall code.  (BTW, are there any
know issues with
>> > >  > verifier
>> > >  > >  > that
>> > >  > >  > > > > needs to be fixed?)
>> > >  > >  > > > >
>> > >  > >  > > > > Having said that I thought that we are
testing up to 6
>> > >  > snapshots per
>> > >  > >  > > > > week so we may pick up any interim snapshot
that has
>> > everything
>> > >  > >  > > > > required and shows good testing results,
make it
>> > 'official' -
>> > >  > i.e.
>> > >  > >  > > > > publish it ... with proper label - M5.5_Eclipse
or
>> > something
>> > >  > else to
>> > >  > >  > > > > avoid confusions and to state clearly that
the release
>> > it
>> > >  > targeted
>> > >  > >  > to
>> > >  > >  > > > > the Eclipse TPTP release.
>> > >  > >  > > > >
>> > >  > >  > > > > Does it make sense and works for all parties?
>> > >  > >  > > > >
>> > >  > >  > > > > The only issue that still unclear for me
is ABI
>> > requirements:
>> > >  > has
>> > >  > >  > the
>> > >  > >  > > > > Harmony team build/test the code to satisfy
ABI or you
>> > can do
>> > >  > it?
>> > >  > >  > > > > (Alexey Petrenko asked this before but
I don't see any
>> > answer)
>> > >  > >  > > >
>> > >  > >  > > >
>> > >  > >  > > > I suppose we can do it, but it should be in
the released
>> > package
>> > >  > too.
>> > >  > >  > If
>> > >  > >  > > > we're going to share building of the module
how it will
>> > looks
>> > >  > like?
>> > >  > >  > > > 1. you'll give us revision
>> > >  > >  > > > 2. and we'll return the compiled libraries
>> > >  > >  > > >
>> > >  > >  > > > or some other way?
>> > >  > >  > > >
>> > >  > >  > >
>> > >  > >  > > OK as I see from your answer - the Harmony team has
to build
>> > >  > binaries
>> > >  > >  > > that satisfy ABI (because we publish binaries that
are
>> > created only
>> > >  > by
>> > >  > >  > > Harmony committers.)
>> > >  > >  > >
>> > >  > >  > > -Stepan.
>> > >  > >  > >
>> > >  > >  > > <SNIP>
>> > >  > >  > >
>> > >  > >  >
>> > >  > >  >
>> > >  > >  >
>> > >  > >  > --
>> > >  > >  > --vvl
>> > >  > >  >
>> > >  > >
>> > >  >
>> > >  >
>> > >  >
>> > >  > --
>> > >  > With best regards,
>> > >  > Alexei
>> > >  >
>> > >
>> > >
>> > >
>> > >  --
>> > >  --vvl
>> > >
>> >
>> >
>> >
>> > --
>> > With best regards,
>> > Alexei
>> >
>>
>>
>>
>> --
>> --vvl
>
>
>
>
> --
> --vvl
>



-- 
With best regards,
Alexei



      ____________________________________________________________________________________
Be a better friend, newshound, and 
know-it-all with Yahoo! Mobile.  Try it now.  http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ
Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message