[aem-devel] Re: AEM work

[Frederic R. (LMC) <Fre...@er...>]

Ardelle,

 As I already told you, I really appreciate your help. AEM is a
great project to work on because it is new. Also it is more than a
notification
mech. AEM brings a new software development model into Linux. For that
reason there is not much documentation to rely on. Which also means that
I want to be very careful regarding what is to be integrated or not. I wi=
ll
also be very carefull not to integrate new features before the current
release
is not completly stable or at least free of bad bugs. Otherwise it will
become
a nightmare to debug.

SMP is not a feature. It is really a needed architecture. Some bad bugs c=
an
make AEM
crash or freeze. This is always the case for any software however. The ga=
me
is
to catch them all and correct them. But I want to be clear on that, when =
I
say
AEM needs to be stabilized it doesn't mean it crashes all the time, I mea=
n
it needs
to be stabilized under high stress or for other sporadic cases. This is
especially
true for SMP. These bugs and pbms are really difficult to correct.

Regarding the test case with 0 as a parameter, this is clearly a paramete=
r
missuse,
Alhough the input value is not tested correctly, this is not the problem.

The only way to get hands dirty is really to use it a lot. Is it the only
test you did ?
How much timers are you able to start ? Did you try other examples ?
a TCP/IP server for example ? Try to run different configurations and the=
n
stress
them.


Frederic

----- Original Message -----
From: "Fan, Ardelle" <ard...@in...>
To: "Frederic Rossi (LMC)" <Fre...@er...>
Cc: "Hu, Boris" <bor...@in...>; <aem...@li...>
Sent: Monday, June 23, 2003 4:46 AM
Subject: RE: AEM work


Hi Frederic

Maybe we were too hurried. We have already downloaded the source code and
went through roughly.
And now we are setting up the environment and trying all test cases.

As for our intention about AEM, to be frankly, we just hope to contribute.
We really wish to do some help to AEM project.
UP/SMP issue is a good point and we are investigating from test case to
reproduce the problem. Could you tell us how
did you find that it has this kind of issue? running some test case or ju=
st
leave it with some days to see it hang (or crash)?
We plan to first reproduce it and try to shorten the time before hang
happens in order to ease future debugging.
How do you think about it? And any suggestion about it?

And here is another thing we found in our try: In
aem-testsuite-v0.3/Timer_Pkg, when we pass the command "./timer 0", 5 tim=
er
is created.
And then there are several lines of timeout messages showing up the scree=
n.
After that the linux box(SMP) hung in about ten seconds. I know
that 0 is not a valid parameter for the timer interval. I just want to se=
e
what will happen in such edge condition. Can we take it as the
symptom of the UP/SMP issue?

In last mail I have asked the TODO list, because SMP issue can be a globa=
l
and deep hidden bug. Maybe some small feature
support can be some threads for us to get hands dirty. If you think there
isn't any available for us, just forget about it.

Thank you
Regards/Ardelle


-----Original Message-----
From: Frederic Rossi (LMC) [mailto:Fre...@er...]
Sent: 2003=C4=EA6=D4=C221=C8=D5 2:10
To: Fan, Ardelle
Cc: Hu, Boris; aem...@li...
Subject: Re: AEM work



It's not the way it work. If you plan to contribute
I firmly encourage you to first download the source code,
try it out, and get the exact idea of what we
are trying to acheive. You can use this mailing list
to send patches or communicate, it's a good idea.


You want to use AEM to do what exactly?
I you need this type of mechanism, you probably know
better than me which feature to implement.

But if you just want to contribute, once again, I have no
problem with, this but you have to show you understood
what AEM does first.


F.




----- Original Message -----
From: "Fan, Ardelle" <ard...@in...>
To: "Frederic Rossi (LMC)" <Fre...@er...>
Cc: "Hu, Boris" <bor...@in...>; <aem...@li...>
Sent: Friday, June 20, 2003 4:18 AM
Subject: AEM work


Hi Frederic

Could you please add me and Boris to the developer list? Thank you.
My sf account is ardelle_fan.
Boris's is boris_hu.
We will setup the environment and run the testsuite to have a taste first.
:)

From cgl-discussion mailing list, there is several line in PoC Meeting
Minutes 6/19/03 that:
"
......
Opens
-----
AEM to be submitted to lkml.
......
"
Have you seen that? Could I know your plan of AEM submission to lkml?

And I think we'd better settle down a TODO list for several monthes. So t=
hat
we can read
the source with focus and plan our long term development tasks.
According to your mail, it seems that SMP stabilization is a feature we
could focus on.

How do you estimate the gap between current 2.5 implementation and the
standard of being accepted by lkml? I am considering that it is also a go=
od
point for us.

And when I read through the "OSDL CGL Requirements for Low-Level
Asynchronous Events", there are a lot of Priority 2 features required.
According to that document, p2 is for CGL3.0. But on the base of a stable
AEM on both UP and SMP, and also acceptable to lkml, could we choose some
imortant features implemented in CGL2.0? One more question, is all P1
features implemented in current 2.5.60 patch?

Sorry for so many questions and thanks for your help!

Ardelle Fan
Software Engineer,  Intel Corporation
ard...@in...

These are my opinions and absolutely not official opinions of Intel Corp.