**** BEGIN LOGGING AT Sun Feb 11 13:17:12 2001

13:17:12 -->	RobertFitzsimons (bob@194.125.56.12) has joined #jos
13:17:12 ---	Topic for #jos is JOS - The Java Operating System (http://www.jos.org)
13:17:12 ---	Topic for #jos set by jewel at Sun Feb 11 10:28:44 2001
13:17:27 <RobertFitzsimons>	Hi John.
14:01:26 <jewel_>	hey
14:03:38 <RobertFitzsimons>	How's it going?
14:04:05 <jewel_>	Quite well, I'm glad you're here I need some help with all the locking stuff
14:04:26 <jewel_>	I hope all is well with you?
14:05:27 <RobertFitzsimons>	Yeah, I'm ok.
14:06:29 <jewel_>	Are you at work now?
14:06:55 <RobertFitzsimons>	Yes.
14:07:55 <jewel_>	I've been writing a lock suitable for the kernel build of teaseme
14:08:11 <jewel_>	It's a recursive mutex that depends on a simpler lock (I'm thinking kspinlock)
14:11:52 <RobertFitzsimons>	Ok, what is this lock going to be used for in the teaseme?
14:12:36 <jewel_>	The lock I'm writing will be used to implement Java monitors
14:12:46 <jewel_>	It will also be used to lock internal teaseme structures
14:13:52 <jewel_>	It will also require a mechanism to put a thread to sleep and reawaken it
14:16:29 <RobertFitzsimons>	I did write kmutex_lock and kmutex_unlock already which will yield if it can't get the lock.
14:16:58 <jewel_>	Ah, excellent, so I don't have to implement that
14:16:58 <RobertFitzsimons>	kthread_sleep doesn't work yet.
14:17:52 <jewel_>	Can you send me the code for that ?
14:18:18 <RobertFitzsimons>	Yes, I'll send it now.
14:19:52 <jewel_>	Thanks
14:26:34 <RobertFitzsimons>	Ok, I've sent it now.
14:26:50 <jewel_>	ok, got it
14:28:58 <RobertFitzsimons>	It shouldn't be too long before dynamic loading of a JVM is supported.  I changed startup code to allow the loading of Grub modules but I can't jump to them yet.
14:31:10 <jewel_>	Cool
14:31:32 <jewel_>	How does that work, where does grub put everything in memory?
14:32:59 <RobertFitzsimons>	Grub gives a pointer to a data structure which contains the location, all I do with that information is set the memory as allocated so the VMM wont over write it.
14:33:19 <RobertFitzsimons>	The code is in init.c.
14:33:54 <jewel_>	So one pointer per module?
14:36:10 <RobertFitzsimons>	Basically, the hard part will be figuring out which is the correct module to load, I will proably do something like grub does (have a header in the module).
14:38:08 <jewel_>	And then how does the code for the module get it's jump targets adjusted / link to external symbols?
14:39:14 <RobertFitzsimons>	I'm sure yet, I'm going to look into it.
14:42:22 <jewel_>	I need to borrow some the RJK code for the linux kernel bulid
14:42:46 <jewel_>	(The atomic routines etc)
14:43:13 <jewel_>	Do you think I should just copy it, or should we build a little library?
14:46:40 <RobertFitzsimons>	Which every way you want to do it, the code will be release under the GPL license.
14:47:54 <jewel_>	of course, I've noticed that a lot of projects put a header at the top of each source file specifying what the license is
14:48:04 <jewel_>	I think we should do this for RJK / teaseme etc
14:48:44 <RobertFitzsimons>	Yes, that's the plan for RJK.
14:49:28 <RobertFitzsimons>	Before I release I'm going to get my employer to sign a release on my code first.
14:51:11 <jewel_>	Smart move
14:51:53 <RobertFitzsimons>	I just have to get around to doing something about it. ;)
14:53:28 <jewel_>	Is your employer well-disposed to such things? (You writing public domain code)
14:56:11 <RobertFitzsimons>	I think so, but I don't think there will be a problem with it.
15:26:48 *	jewel_ reboots the development box
15:35:27 *	RobertFitzsimons goes to get some food.
16:14:02 *	jewel_ is going to leave soon
16:14:33 <RobertFitzsimons>	Is there anything you would like me to work on in RJK?
16:14:53 *	jewel_ thinks
16:15:20 <RobertFitzsimons>	Other then dynamic loading.
16:15:34 <jewel_>	Um, not really
16:16:02 <jewel_>	I don't know if you have, but it would be useful if you could test the locking and threads in RJK before we start porting teaseme
16:17:24 <RobertFitzsimons>	There currently being used by the kernel, but I'll do a bit more testing on them.
16:17:26 <jewel_>	We should also have a chat (including Quinn) about how we're going to feed interrupts to the java code
16:18:11 <RobertFitzsimons>	I think we should use the real time spec for Java for this.
16:18:23 <jewel_>	I'm going to be busy for an hour or two or three, but will come back to fix up the kernel build (segfaulting)
16:18:28 <jewel_>	what is the real-time spe?
16:19:43 <RobertFitzsimons>	I belive Sun has a real-time spec for applications that require it, I don't know where the spec is I'll have to look for it.
16:20:29 <jewel_>	ok
16:20:33 <jewel_>	I'm looking on google
16:22:15 <jewel_>	http://java.sun.com/aboutJava/communityprocess/jsr/jsr_001_real_time.html
16:22:23 <RobertFitzsimons>	Yes, that's the one.
16:26:48 <jewel_>	http://www.javaseries.com/rtj.pdf
16:28:27 <RobertFitzsimons>	I have to download a pdf reader for Linux 5.5Mb. :(
16:28:47 <jewel_>	Don't you have gv?
16:30:07 <RobertFitzsimons>	Doesn't look like it.
16:30:19 <RobertFitzsimons>	... like I have it...
16:30:48 <jewel_>	I was reading some news articles aobut the playstation 2 and 3 today
16:31:06 <jewel_>	i was thinking that a port of JOS to a device like that would be really really cool
16:32:00 <RobertFitzsimons>	Yeah, I was thinking of porting RJK to my Agenda PDA next. :)
16:33:10 <jewel_>	what is that?
16:33:48 <RobertFitzsimons>	http://www.agendacomputing.com/
16:34:44 <jewel_>	I tried evolution for the first time today
16:35:58 <jewel_>	what is this agenda thing, what CPU does it have? Does it have infrared ?
16:37:51 <RobertFitzsimons>	It's a Linux Personal Digital Assistant with a MIPS processor, 16 Mb flash, 8 Mb Ram.  Yes it's got infrared.
16:38:50 <jewel_>	Does gcc compile to the NEC chipn?
16:39:16 <jewel_>	what does it cost?
16:39:39 <RobertFitzsimons>	Yes, the GNU tools work, gcc, bintuils, etc.
16:40:10 <RobertFitzsimons>	You can get a dev unit for about 180 USD.
16:40:24 <jewel_>	have you tried evolution?
16:40:48 <RobertFitzsimons>	No not yet, is that something to do with Gnome?
16:42:04 <jewel_>	Yes, it's a helix gnome application, it's basically what Outlook is
16:42:16 <jewel_>	Does email, contacts, calendar, scheduling etc
16:42:47 <jewel_>	But how would you run new code on the agenda? Does it boot off flash rom?
16:43:04 <RobertFitzsimons>	Oh, yes I remember it now, I haven't used it before.
16:43:44 <RobertFitzsimons>	The developer units are only flash ram so you can load you own OS onto it.
16:43:54 <jewel_>	super cool
16:44:06 <jewel_>	the other's are ROM? they're cheaper?
16:44:41 <RobertFitzsimons>	No ROM units yet, but they should be cheaper.
16:45:56 <jewel_>	Can you view that pdf?
16:46:03 <jewel_>	How do you like it?
16:48:20 -->	_Quinn (mysql@165-82-84-50.students.haverford.edu) has joined #jos
16:48:36 <jewel_>	Ah I had to install gs-pdfencrypt and can see it now
16:48:38 <jewel_>	hey _Quinn
16:49:00 <--	_Quinn has quit (Leaving)
16:49:26 -->	_Quinn (mysql@165-82-84-50.students.haverford.edu) has joined #jos
16:49:36 <jewel_>	_Quinn: http://www.javaseries.com/rtj.pdf
16:49:37 <RobertFitzsimons>	Yes, that's why I had to download 5.5 Mb.  I'm now getting gv aswell.
16:49:39 *	_Quinn wonders about ircii flaking out.
16:49:50 *	jewel_ uses irssi
16:49:52 <_Quinn>	jewel -- sorry, missed the first part of that.
16:50:03 <_Quinn>	real-time-java?
16:51:13 <jewel_>	yes
16:51:19 <RobertFitzsimons>	Yes, we might be able to use it for interrupts, etc.
16:51:19 <jewel_>	dead-cat meeting ?
16:51:25 <_Quinn>	ah.
16:51:31 <_Quinn>	dead-cat?
16:52:05 <jewel_>	it's in the doc :-)
16:52:12 <_Quinn>	Ok :)
16:52:23 <jewel_>	It doesn't seem to mention interrupts, let me see about exceptions ...
16:52:39 <_Quinn>	How long will you two be here?  I need to shower and eat soon.
16:52:58 <jewel_>	Yeah check p28
16:53:07 <jewel_>	I'm going to leave soon, maybe be back in 1 or 2 hours
16:53:24 <_Quinn>	OK, that sounds good to me.  I'll try to read the rtj.pdf by then.
16:54:03 <RobertFitzsimons>	I'm going to be here.
16:54:24 <RobertFitzsimons>	.. here for a few more hours.
16:54:56 <jewel_>	Check the raw memory access on p32
16:55:31 <jewel_>	see you guys later
16:55:38 <_Quinn>	OK
16:55:45 <--	_Quinn has quit (Leaving)
16:58:45 <--	jewel_ has quit (Ping timeout for jewel_[pta-dial-196-31-186-205.mweb.co.za])
17:51:59 -->	_Quinn (tmiller@165-82-84-50.students.haverford.edu) has joined #jos
17:52:11 <RobertFitzsimons>	WB
17:52:20 <_Quinn>	Good to be back :)
17:52:54 <_Quinn>	Except that xpdf doesn't read encrypted PDFs.
17:53:03 <_Quinn>	I think I need a new version...
17:54:27 <RobertFitzsimons>	You need to get acroread or gv with encrypt support.
17:54:51 <RobertFitzsimons>	xpdf doesn't have support for encrypted PDF's.
17:55:35 <_Quinn>	As of 0.91, it's supposed to (according to foolabs.com/xpdf)
17:55:39 <_Quinn>	we'll see :)
17:56:01 <RobertFitzsimons>	Ok, maybe. :)
17:58:51 *	_Quinn goes merrily compiling
18:13:53 <_Quinn>	I'll bbiab.
18:13:54 <--	_Quinn has quit (Leaving)
18:36:27 -->	jewel (jewel@pta-dial-196-31-185-165.mweb.co.za) has joined #jos
18:36:57 <jewel>	Hi
18:37:07 <jewel>	is todd gone?
18:38:16 <RobertFitzsimons>	He was back in about half an hour ago, he said he'd be back "<_Quinn> I'll bbiab.".
18:38:40 <jewel>	What do you use your agenda for?
18:39:43 <RobertFitzsimons>	At the moment not much, it's still in beta stage.
18:40:03 <RobertFitzsimons>	I've got some notes and stuff in it.
18:40:20 <jewel>	But how do you enter any data without a keyboard?
18:42:24 <RobertFitzsimons>	It's got hand writing recognition software, and an on-screen keyboard.
18:45:37 <jewel>	wow, is it easy to enter data?
18:46:32 <RobertFitzsimons>	It's not that easy, but there is work being done to improve it.
18:47:16 <jewel>	Cool, and what desktop software does it interface with?
18:49:33 <RobertFitzsimons>	None yet, but there are plans for outlook support, etc.  Also you can use ppp and rsync if you like.
18:52:16 <jewel>	But then you can only copy raw files
18:53:07 <RobertFitzsimons>	Yes, for the moment.
18:53:39 <jewel>	Cool, I suppose it's easiest to buy in the US
18:54:25 <jewel>	Do you think we could run a JVM in 8MB?
18:54:33 <RobertFitzsimons>	That were I got it but you can get it through Europe.
18:55:29 <jewel>	The Sony emotion-2 has something like 256MB dram on the chip
18:55:40 <RobertFitzsimons>	Sure, 8Mb no problem if you cut down on the class libraries.
18:56:57 <jewel>	We should think about it, we already have the linux kernel module, they've written a graphical library
18:57:21 <jewel>	we still need to write code to access the network interfaces though
18:59:06 <RobertFitzsimons>	Current teaseme might be a bit tight (remember my tests with the 8MB laptop), but we can work on it.
19:01:48 <jewel>	Yep, but currently we have little control over memory
19:02:06 <jewel>	The heap has to be about 1MB, although we could live with 0.5 MB
19:02:22 <jewel>	And the class structures just eat kernel memory and never get free'ed
19:02:28 <jewel>	I assume there's no swap on the device?
19:03:51 <RobertFitzsimons>	No, and there is only about 3MB ram free in normal mode.  But if we had a custom kernel, etc we could get more.
19:26:19 -->	_Quinn (tmiller@165-82-84-50.students.haverford.edu) has joined #jos
19:26:44 <RobertFitzsimons>	Wb again. ;)
19:26:45 <_Quinn>	Sorry I took so long, guys, but I accidentally nuked X while diddling around with truetype fonts for xpdf.
19:27:02 <jewel>	wb _Quinn
19:27:07 <jewel>	hehe
19:27:12 <jewel>	why don't you use gv?
19:27:56 <_Quinn>	because gv bombs on encrypted pdfs also :)
19:28:31 <jewel>	not if you install the other stuff (there is a debian pkg for it)
19:28:37 <_Quinn>	shrug :)
19:30:00 <jewel>	RobertFitzsimons and I were discussing his little Agenda PDA and the possibilities of running Java on it
19:30:29 *	_Quinn nods
19:30:58 <jewel>	I think with 16MB it would be cool, are there any plans to expand it?
19:31:58 <RobertFitzsimons>	Actualy the best (geek) thing about it is that it is a really linux system, you can telnet into it, nfs mount, remote X session, etc.
19:32:29 <jewel>	yeah, I see it can do routing, ppp and all that stuff too
19:32:44 <RobertFitzsimons>	I don't think there are any plan for the VR3 version, but the next one should have more.
19:44:23 <_Quinn>	jewel, I was wondering about classpath and teaseme
19:44:36 <_Quinn>	How much of classpath did you have to change to get it to work?
19:45:28 <_Quinn>	We don't really have the manpower to maintain our own copy of it, and merging their changes in could be painful...
19:45:44 <_Quinn>	You did your work on the native side of things?
19:47:45 <jewel>	To get classpath to work with kissme required quite a few changes
19:48:06 <jewel>	There was a problem with their exception-handling and trace mechanisms which i had to change
19:48:20 <jewel>	But this has change in classpath 0.02, so I don't know if it will be easier now
19:48:58 <jewel>	On the native side it's pretty much unchanged for kissme, but for teaseme I had to rip out all the user-side code
19:49:10 <_Quinn>	Well, I guess we don't have to worry about upgrading until we find a problem with the version we've got now; I was just recalling the vm/ directory in classpath --
19:49:27 <_Quinn>	-- how much that could cover of what needs to happen.
19:49:39 <jewel>	The changes aren't that dramatic, they only affect a few files
19:49:51 <jewel>	The problem is that classpath keeps adding native methods to some of the core classes
19:49:59 <_Quinn>	ah :)
19:50:11 <jewel>	Then I have to implement these to catch up, which is why I haven't kept current with the CP releases
19:50:58 <_Quinn>	Could you -- this is speculative question -- wrap the equivalent kernel calls in user-space replacements and compile classpath's tree w/o -lc?
19:51:05 <jewel>	But they were probably just missing from earlier core classes, and are part of the JDK (or a new version of the JDK)
19:51:37 <jewel>	No, you can't make kernel calls from within the kernel
19:51:57 <jewel>	Many of the structures used by libc don't exist in kernel-space
19:52:51 <_Quinn>	OK.  I guess I'll just have to compare the classpath version and your version and see what I'm not getting.
19:53:04 <jewel>	Oh you mean the other way round?
19:53:55 <jewel>	Basically, when classpath does things like fopen, or even open(), fread, read, write etc
19:54:11 <jewel>	we have to make a fairly complex wrapper which uses kernel structures
19:54:18 <_Quinn>	I was thinking about the way JJOS/decaf worked, which was to provide a 'fake' libc that used 'kernel' calls/code.
19:54:27 <jewel>	this means changing a fair amount of the classpath native code
19:55:31 <_Quinn>	I was thinking maybe you could move all those wrappers out of the classpath tree, so we can just automagically suck up their updates.
19:55:55 <_Quinn>	But I don't which one is easier -- doing the updates by hand, or moving the wrappers outside (maintaining a subset of libc).
19:56:18 <_Quinn>	But like I said earlier, this isn't a big issue for now
19:56:35 <RobertFitzsimons>	A lot Classpath isn't going to work very well when we have a native build, i.e. when we have our own kernel, jvm, drivers.
19:56:46 -->	jewel_ (jewel@pta-dial-196-31-185-165.mweb.co.za) has joined #jos
19:57:14 *	_Quinn shrugs
19:57:22 <jewel_>	server dropped me
19:57:24 <RobertFitzsimons>	_Quinn> I was thinking maybe you could move all those wrappers out of the classpath tree, so we can just automagically suck up their updates.
19:57:24 <RobertFitzsimons>	<_Quinn> But I don't which one is easier -- doing the updates by hand, or moving the wrappers outside (maintaining a subset of libc).
19:57:24 <RobertFitzsimons>	<_Quinn> But like I said earlier, this isn't a big issue for now
19:57:24 <RobertFitzsimons>	<RobertFitzsimons> A lot Classpath isn't going to work very well when we have a native build, i.e. when we have our own kernel, jvm, drivers.
19:57:41 <_Quinn>	I really don't know; I haven't looked closely enough at classpath.
19:57:42 <--	jewel has quit (Ping timeout for jewel[pta-dial-196-31-185-165.mweb.co.za])
19:58:18 <_Quinn>	It's just that a class library is a big enough project that even GNU is having a hard time with it, so we certianly don't have the manpower.
19:58:28 <jewel_>	what is significant is that there is as little native code in classpath as possible
19:58:46 <RobertFitzsimons>	At least the pure Java part should be ok (I hope).
19:58:47 <jewel_>	we also only depend on a subset of that (java.io / java.lang)
19:59:14 <_Quinn>	jewel -- at some point, we'll want everything that's not pure java.
19:59:22 <jewel_>	later we will have higher level java code doing the equivalent of what the java.io native code does now (ie open files, read/write to them, seek them etc)
19:59:22 <_Quinn>	(the pure java bits people can get from sun, et. al.)
20:00:09 <jewel_>	so I think it would be a waste to spend effort on integrating with that native code when ultimately we're not going to use it
20:00:27 <_Quinn>	duh!
20:00:32 *	_Quinn slaps himself in the head
20:00:38 <_Quinn>	OK, I get what you're talking about now.
20:01:02 <_Quinn>	We'll be rewriting everything below the class library level anyway, right...
20:01:19 <jewel_>	I was reading a book I have the other day, Practical file system design with BeFS
20:01:31 <jewel_>	Have you guys heard/read anything about BeFS?
20:01:49 <_Quinn>	... dispatching native calls from that library via BCNI, I guess...
20:01:51 <_Quinn>	Nope.
20:03:52 <jewel_>	Well it's a modern FS with some cool concepts
20:03:52 <jewel_>	It's a journalling fs, they call it 64-bit because it supports huge files and partitions
20:03:52 <jewel_>	In addition to that it supports extended attributes
20:03:52 <jewel_>	This makes it more like a database than the traditional FS'es
20:04:07 -->	jewel (jewel@pta-dial-196-31-185-165.mweb.co.za) has joined #jos
20:04:25 <jewel>	You can build indices on these attributes, making for very fast querying of info about files
20:04:45 <jewel>	it allows applications to use the FS for storage instead of some other mechanism
20:05:04 <jewel>	For example email can be stored in the FS, with attributes for the From, To, Subject headers etc
20:05:32 <jewel>	You can also establish persistent queries which are updated whenever the data they are watching changes
20:06:12 <_Quinn>	Interesting...
20:06:22 <jewel>	Anyway, I was reading about the design of the v-nodes and the way different filesystems are managed in one namespace
20:06:44 <_Quinn>	BeFS for linux? :)
20:06:52 <--	jewel_ has quit (Ping timeout for jewel_[pta-dial-196-31-185-165.mweb.co.za])
20:07:03 <jewel>	Which is why I say that we won't be needing the java.io native code eventually
20:08:04 <jewel>	There is some befs attempt for linux
20:08:10 <jewel>	Don't know how far it got
20:08:33 <_Quinn>	right.
20:13:34 <RobertFitzsimons>	We should just implement ext2 and fat first and let some else do the rest, we can get around to the other when we have a full running system.
20:13:47 <_Quinn>	Yup.
20:14:01 <jewel>	of course
20:14:12 <jewel>	even implementing those two is a massive task
20:14:45 <jewel>	I think we could write the java VFS structure, but just have the native linux implementation underneath 
20:14:48 <jewel>	as a starting point
20:14:58 <_Quinn>	Of course, we should sit on filesystem drivers for a while yet.
20:15:07 <_Quinn>	(In some fashion or another.)
20:15:20 <_Quinn>	Until we'll sure that our architecture is ready.
20:15:41 <_Quinn>	or s/architecture/design, s/ready/good
20:16:25 <jewel>	What does this error mean: storage class specified for parameter `kint'
20:16:48 <RobertFitzsimons>	What we should be thinking about is do we want to continue with teaseme on Linux or teaseme on RJK?
20:17:17 <jewel>	Personally I think both
20:17:38 <RobertFitzsimons>	The type kint isn't defined, look in kimplementation.h.
20:18:16 <RobertFitzsimons>	I was thinking to save effort just port the kinterface to the Linux kernel.
20:18:16 <jewel>	doesn't kinterface include kimplementation.h ?
20:18:27 <jewel>	yes, I was thinking the same
20:18:31 <_Quinn>	Having another kernel couldn't hurt, but I would prefer it if teaseme on Linux weren't 'held back' by compabality with RJK.
20:19:01 <RobertFitzsimons>	It should, is that the only error?
20:20:26 <jewel>	I don't think it will be held back
20:20:41 <jewel>	Rather some features just won't be available on RJK (as it is now with user-mode)
20:20:42 <RobertFitzsimons>	RJK should have most if not all the features required within the next few weeks, the only disadvanage I can see having to reboot instead of insmod/rmmod.
20:20:57 <jewel>	RobertFitzsimons: there is another error above which I'm looking at now
20:21:19 <jewel>	RobertFitzsimons: and the FS support, although we don't write to the FS yet in teaseme
20:21:54 <RobertFitzsimons>	Quinn, would you like a copy of the lastest version of RJK?
20:23:14 <_Quinn>	Sure.
20:23:43 <_Quinn>	Well, having to reboot and not having any fs write abilites could be an advantage for certain kinds of debugging :)
20:24:00 <RobertFitzsimons>	But we can have Grub load a FS type image into memory that we can use.
20:25:56 <jewel>	Or write a Java memfs or something
20:26:37 <RobertFitzsimons>	I find the best way to do testing with RJK is with bochs, read the kernel from a floppy image and setting break points etc.
20:29:49 <jewel>	Building gcc  -c -D__KERNEL__ -DMODULE -O -Wall -DKISSME_LINUX_KERNEL  -c -I/usr/src/linux/include  -I. -Ivm ./vm/sys_linux_kernel/locks.c -o ./vm/sys_linux_kernel/locks.o
20:29:51 <jewel>	In file included from kinterface.h:8,
20:29:53 <jewel>	                 from ./vm/sys_linux_kernel/locks.c:7:
20:29:56 <jewel>	kimplementation.h: In function `sys_mutex_create_inplace':
20:29:57 <jewel>	kimplementation.h:12: storage class specified for parameter `kint'
20:32:58 <RobertFitzsimons>	I email the source to you Quinn.
20:33:28 *	_Quinn scurries over to his email
20:33:30 <jewel>	why am I getting the error?
20:33:39 <RobertFitzsimons>	I'm not sure.
20:34:51 <jewel>	I'm confused by "storage class specified"
20:35:19 <RobertFitzsimons>	Is there something wrong with this "typedef signed long int kint;"?
20:36:38 <_Quinn>	OK, well egcs likes it just fine.
20:37:17 <jewel>	yes it worked with me before (and I can compile RJK standalone no problem)
20:37:25 <_Quinn>	And so does gcc-2.95.2
20:37:54 <jewel>	i wonder what is conflicting ...
20:37:54 <_Quinn>	Oh... how do I do the combo-build/
20:38:04 <_Quinn>	er, s\/\?\
20:40:37 <RobertFitzsimons>	Is that vaild C or C++ code?
20:41:26 <jewel>	The "typedef ... " ?
20:41:29 <RobertFitzsimons>	Yes.
20:41:31 <jewel>	or my code ?
20:41:36 <jewel>	Yes it is
20:42:32 <jewel>	Ok I'm compiling successfully now
20:42:43 <_Quinn>	Is it a single floppy I'm supposed to be making for RJK?
20:43:20 <RobertFitzsimons>	Yes.
20:43:48 <RobertFitzsimons>	It's just a fat formated/grub bootable disk image.
20:44:19 <RobertFitzsimons>	Jewel, what did you change?
20:44:23 <_Quinn>	oh!
20:44:31 <_Quinn>	So I un-gz first, then?
20:44:36 <RobertFitzsimons>	Yes.
20:44:40 <_Quinn>	OK, thanks :)
20:45:48 <jewel>	I was trying to declare a structure that had a kuint in it
20:46:05 <jewel>	And was then using that structure in prototypes for functions in headers
20:46:10 <jewel>	Somehow this caused a lot of problems
20:46:51 <RobertFitzsimons>	Maybe you were using in before it was declared?
20:48:28 <_Quinn>	RF: what does the boot floppy do (now that I've got set up)?
20:48:56 <jewel>	RobertFitzsimons: I think i was, the header dependencies were quite complex
20:51:08 *	jewel bounces the box
20:51:37 <RobertFitzsimons>	Grub will load a file called kernel.elf, then the kernel will set it's self, up start two threads which will print some text, and loop forever.
20:51:48 <_Quinn>	OK.  Did you want me to try it out?
20:52:42 <RobertFitzsimons>	It's up to you.  If you have a second computer give it a good, if not you should problaly leave it until later.
20:53:34 <RobertFitzsimons>	I've tried it on a few different types of computer without any problems.
20:55:23 <_Quinn>	Well, it worked on my laptop :)
20:59:23 *	jewel builds teaseme
21:04:17 *	jewel builds teaseme again
21:10:33 *	jewel bounces the box
21:14:08 <_Quinn>	Do either of you know much about HTTP?
21:14:25 <jewel>	a little bit
21:14:37 *	jewel blasphemes when he discovers his favourite tea has run out
21:15:00 <RobertFitzsimons>	A bit, I've read the RFC docs, etc.
21:15:01 <_Quinn>	the 'Location' header will cause the browser to stop loading the current page and get the new one, right?
21:15:41 <jewel>	Urm, I remember something like that, it has a timeout too
21:16:00 <jewel>	But does that have to do with http or html?
21:16:49 <RobertFitzsimons>	There is something like that in html, but I'm not sure about http.
21:16:53 <_Quinn>	OK.  Because I'm having this problem with authenticating a user with PHP -- I use the location header to redirect after they've posted a valid username/password, but the page that I direct them to won't recognize the stored auth. until it's reloaded.  Which is just bizarre.
21:17:32 <_Quinn>	Maybe it's legal not to propogate stuff atfter the '?'?
21:18:51 <RobertFitzsimons>	I'm not sure.
21:19:02 <_Quinn>	Hmmm....
21:19:16 *	_Quinn will ponder this
21:23:31 *	_Quinn has other things to do.  see you around.
21:23:33 <--	_Quinn has quit (Leaving)
21:26:54 *	jewel traces the segfault
21:37:53 *	jewel realises how stupid C compilers are
21:42:28 *	jewel performs self-flagellation when he realises what a silly mistake he's made
21:43:36 <RobertFitzsimons>	me learn't during the week that the problem is almost always with his new code.
21:45:01 <jewel>	yes, in this case too
21:45:13 <jewel>	but I copied the same code into two places and only debugged the first piece
21:46:16 <jewel>	What is the maximum pid in linux?
21:47:29 <RobertFitzsimons>	I think it's 32768.
21:48:03 <RobertFitzsimons>	It was only changed recently to use the full 32 bits.
21:49:20 <jewel>	Hah! No segfault
21:52:26 <jewel>	I was masking out the high two bits of the pid number in the thin lock word
21:52:37 <jewel>	The pid was 11000 something, so it was getting modified
21:55:14 <jewel>	but there's still another bug
21:55:19 <jewel>	I'll leave it for tomorrow
21:55:30 <jewel>	anything else you want to discuss before I go?
21:55:39 <RobertFitzsimons>	No not at the moment.
21:57:09 <jewel>	Ok, I borrowed another monitor to replace the one that died on my other box, so I'll try compile some stuff with RJK and see what happens
21:57:53 <RobertFitzsimons>	Yeah, I hope to work on dynamic loading over the next week.
21:58:17 <jewel>	It will be very exciting if we can get teaseme booting on RJK
21:59:14 <RobertFitzsimons>	It would boot using three different systems, Linux host, Linux kernel and RJK, that would be cool.
21:59:46 <jewel>	yes, it gives a lot of scope for debugging (different environments == stressing the code in different ways)
22:00:35 <jewel>	and allows us to play with a variety of different things (RJK, raw hardware access, linux - already has routing and network stuff, userspace - good debugging and stable file and network support)
22:02:26 <RobertFitzsimons>	Yes, it would be great.
22:02:42 <jewel>	ok, see you later
22:02:46 <RobertFitzsimons>	Bye.
22:02:57 <--	jewel has quit (sleeping)
22:37:13 ---	You have left channel #jos
**** ENDING LOGGING AT Sun Feb 11 22:37:13 2001