Its been brought to my attention the
base team has "closed" the previously suggested bug,
having collected plenty of information
for their M6 purposes, and have no automated way to
process it .... so, I've created a bug
in WTP to attach performance logs to, if you collect them
during M4 testing.
Again, this isn't so much to find/fix
problems (yet), but more to see if this instrumentation seems to
provide useful information.
===== Original posting =====
David M Williams/Raleigh/IBM@IBMUS
23/04/2005 01:56 AM
Suggest testing with performance
test flag "on"
As teams do "top to bottom"
tests this week in preparation for our WTP M4,
I have a suggestion that can help the
base Eclipse team collect performance
data (and in turn, help them help us :)
With a fresh install of M6, there's
a default .options file that begins as below.
Options for all performance debug flags in the platform
Captures performance event information (See org.eclipse.core.runtime.PerformanceStats)
Assuming this default options file,
all you need to do is run eclipse with the -debug option on the command
This will then collect a file full of
things that appear to be taking longer than usual
(builds, view refreshes, opening (some)
This file is kept in the .metadata directory
of your workspace you are working with.
(You might want to start with a fresh
copy every now and then ... it doesn't take long
for them to get pretty big, depending
on what you are doing).
Then attach that file to https://bugs.eclipse.org/bugs/show_bug.cgi?id=88096
with some description of your machine's
speed, what you were doing (e.g. "webtools,
creating EJB sample" ... or similar.
I believe the base team is still evaluating
this "instrumented" technology, to see
if its helpful in narrowing down broad
areas to investigate for improvements, but
suspect this may be one of the first
opportunities to collect the data with a substantial
extension to the base platform.
Note: I'm saying all this as a kind
of low priority thing, and mention it only because
its the kind of thing that should just
take you an extra 10 or 15 minutes to do ...
I'm NOT asking you diagnose, fix, or
instrument any of our own code (yet :)
The purpose is just to see if running
"webtools" shows a much different pattern than running