|Greg Lewis||Jul 25, 2002 11:20 am|
|Bill Huey (Hui)||Jul 25, 2002 2:42 pm|
|Greg Lewis||Jul 25, 2002 5:36 pm|
|daichi||Jul 25, 2002 8:04 pm|
|Greg Lewis||Jul 25, 2002 8:12 pm|
|Greg Lewis||Jul 25, 2002 9:58 pm|
|Bill Huey (Hui)||Jul 25, 2002 10:04 pm|
|daichi||Jul 25, 2002 10:49 pm|
|Dev Zero G Ltd||Jul 26, 2002 6:33 am|
|Greg Lewis||Jul 26, 2002 7:21 am|
|Greg Lewis||Jul 27, 2002 5:27 pm|
|Bill Huey (Hui)||Jul 27, 2002 7:33 pm|
|Subject:||Re: 1.3.1 patchset 7|
|From:||Bill Huey (Hui) (bil...@gnuppy.monkey.org)|
|Date:||Jul 25, 2002 10:04:48 pm|
On Fri, Jul 26, 2002 at 10:06:36AM +0930, Greg Lewis wrote:
I thought there was some contention that -STABLE's pthreads weren't up to the task either, or is that not correct?
Yes, it's got a bug suspension bug that's fixed in -current. Then I found/fixed a bug in -current concerning suspension at thread creation time.
Also, how do you feel about writing a paragraph or so on building HotSpot? I'm sure it would help a lot of people :).
Sure, it's pretty easy if your patch is using the sources I commited recently. You basically need a fully functioning -current. That might be a bit difficult since it's going through a lot of volatile changes.
Here's what you do:
1) Install gcc31. That comes with -current, but you might want to build it from the ports anyways since the "platform_i486" expects that.
2a) Move into "hotspot1.3.1/build/linux/"
b) Use: setenv ALT_BOOTDIR /usr/local/linux-jdk1.3.1/
# if you have an SMP machine otherwise left this line out. setenv HOTSPOT_BUILD_JOBS 4
c) Type "gmake core compiler1 compiler2" to build all six HotSpot versions.
d) Install it using some variation of:
set files="linux_i486_core/product/libjvm.so linux_i486_core/jvmg/libjvm_g.so"
linux_i486_compiler1/jvmg/libjvm_g.so" #set files="linux_i486_compiler2/product/libjvm.so
ls -l $files
rm -rf $destination mkdir $destination
echo cp $files $destination cp $files $destination
ls -l $destination
e) Modify jvm.cfg to use the HotSpot libraries.
f) Run the JVM.
Right now, "core" works best. Compiler[1,2] should work faster, but for some reason it doesn't. I don't know if it's the pthreads library, my nanosleep wrapper or whatever... so that should be checked out by somebody that's got some time and technical ability.
"core" should at least get you a working HotSpot in interpreted mode and use that for now since it seems the fastest. Everything I've tried works well in it so far.
I'm trying to leave the project and do other things outside of the this group right now. I can't keep doing this for free any more since I've got to think of my overall career and facilitate activities that forward it.
HotSpot was the last thing I really wanted to finish. Since I got it largely working now, I've made the effort to focus on other things that are key to what I want to do in the near future.
I was going to do FreeBSD 5.0 hacking, but my experience with the FreeBSD community overall has been pretty negative (not this specific group though) stemming from the days pre-BSDi and now largely because of their clique driven culture. It's a well known problem outline by Matt Dillion. In that regard, this is probably going to be one of the last things I'll be doing for FreeBSD.
We'll see if this is an overreaction, but my faith in FreeBSD overall is pessimistic because of my experiences.
To Unsubscribe: send mail to majo...@FreeBSD.org with "unsubscribe freebsd-java" in the body of the message