The collected works of telackey

I thought I'd share some quick notes on using DINA with Hyper-V.

There may be better ways to do this, but this was my process:

1) Convert the VMDK to a VHD with the freeware tool vmdk2vhd.

2) Resize the drive with the freeware tool VHD Resizer. I used 80GB, but it hardly matters because the extra "space" is an illusion, and will never be used. There is something not quite right about the VHD after the conversion. If you don't also resize the drive, when you try to boot it will complain that wd0 is smaller than its partitions. I expect this is some off-by-one bug, so anything bigger than 60GB will do for the new size.

3) Create a VM in Hyper-V pointing at the VHD. The default settings are fine, except that you don't need nearly 512MB of RAM: 160MB or so is plenty. After it is created, under 'Hardware', remove the standard network adapter and add a legacy network adapter.

4) Start the VM.

5) Hit a key before it boots the kernel and type 'boot -c' to get into userconfig. Then disable ACPI with:

Code: Select all

disable acpi
disable re
quit

If you don't disable ACPI, the network will not function properly and there will be a bunch of warnings about the processor during startup. Unfortunately this needs done every time it boots. :/

6) You may need to update the keyboard settings. They are set by 'encoding' in /etc/wscons.conf. I used a value of 'us'. Restart.

7) The network interface is 'tlp0' under Hyper-V instead of 'pcn0' under VMWare. You'll either need to edit the template files under /usr/dina before configuring the system, or edit /etc/rc.conf afterward to use the right interface.

And that's it! I have no idea if X works under Hyper-V. I've never tried, but I rather doubt it. The important part is at this point you should be able to access it over the network from Windows and SGI machines, and as soon as your IRIX files are copied over, can do netboot installations.

*ADDENDUM*

I found a couple of more things, though not related to Hyper-V specifically.

1) When it rolled over from Feb to Mar, new static BOOTP entries stopped being made properly, and what got entered in dhcpd.conf broke DHCP entirely by setting the word 'from' as the MAC address for the new entry. When the days moved from two digits (29) to one (1), the 'cut' in the script /usr/dina/checkforSGI_BOOTP_reqs was grabbing the wrong field. I haven't really tested it, but my fix was to add:

Code: Select all

sed 's/  */ /g' |
ahead of the 'sort/uniq/cut' part.

2) Every night the cpu got pegged and kept there. I noticed that /usr/dina/startmenu was running even though no users were logged in. A little poking around showed that the daily security checks do something logged in as root, which started startmenu automatically. Because startmenu was trying to read input, it never finished and the security check would spin forever. The fix for me was to edit /etc/security.conf and add this at the bottom:

Code: Select all

check_rootdotfiles=NO
:1600SW: :O2: :O2+: :O2: :1600SW:
Great! I'll fill in the wiki soon. Maybe include some screenshots of the process too.
:1600SW: :O2: :O2+: :O2: :1600SW:
I mentioned this in passing on another post, but I was having trouble with FF loading any largish picture* on a fresh install of 32-bit 6.5.22. Instead of displaying the image, it would give a message like:

Code:
Image "xyz.jpg" cannot be displayed because it contains errors.


I suspected this was an 'rqs' problem (later confirmed with a little work with 'par'). The first time around (hummingbird, first link) a thorough 'rqsall' fixed it. Later I tried to view an even larger picture (Raphael's School of Athens, second link) and it broke again. Trying another complete 'rqsall' didn't fix that.

So I tried this instead, and all seems to be OK:

Code:
#use sh or bash

ldd /usr/nekoware/lib/firefox-bin | grep neko | cut -d'>' -f2- > /tmp/fflibs.txt

#stop and examine the contents of fflibs.txt before running this

for f in `cat /tmp/fflibs.txt`; do
rqs -m -f -l 0xa000000 $f
done


I just typed that so probably worth check for typos. If anyone else hits the same thing, I hope this will help. I had tried requickstarting a library at a time, but gave up and just moved them all as above. For more details on the fundamental problem, try searching for rqsall and 0xa000000. There are some old posts lurking around that gave me the idea to begin with.

*Links:
https://upload.wikimedia.org/wikipedia/ ... 1-edit.jpg
https://upload.wikimedia.org/wikipedia/ ... zio_01.jpg

_________________
:1600SW: :O2: :O2+: :O2: :1600SW:
FWIW, closing and reopening never helped for mine. I imagine it is a question of exactly where the offending libraries have been located in the address space.

_________________
:1600SW: :O2: :O2+: :O2: :1600SW:
Not strictly speaking IRIX/SGI development, but I thought it might be of some interest that CDE was just released open source: http://sourceforge.net/p/cdesktopenv/wiki/Home/

Motif is soon to follow, according to the FAQ.

_________________
:1600SW: :O2: :O2+: :O2: :1600SW:
I gotta know, did you track down all the unaligned memory access? Or did you use 'sysmips(MIPS_FIXADE, 1, 0, 0)' ?

If you could e-mail your patches, I'd appreciate it just out of curiosity. I've had my build working (with Frank's excellent assistance) but hitting BUS errors sitting idle for a few months with a big 'todo' of sitting down and tracking down the errors or recompiling with the FIXADE workaround.

[EDIT, FWIW I was using mips pro, so I'm not sure how similar the problems were with gcc. I see the patch is supposed to be included in the tardist. I'll take a look.]

[EDIT, EDIT: Hmm, I've got it installed and it runs fine, but I don't see the patch file under /usr/nekoware/patch[es] ]
:1600SW: :O2: :O2+: :O2: :1600SW:
I tried wrp running on Ubuntu 14.04 to help out my O2.

It was not very usable. It would seg fault on many sites (eg, amazon.com) or simply not load them (microsoft.com).

It did load sgi.com and google.com, but broke many of the links on the former, and there was no form input for the latter, so no way to search. Not being able to search anything on any site is a biggy.

The idea is clever, but it has a ways to go. If only I had the time, I think I would rewrite it, as I have experience from a very extensive project where I had to dive into the guts of PhantomJS and drive it in odd, interesting, and exhaustive ways, and though it was a close run thing, I lived to tell the tale.

If only...
:1600SW: :O2: :O2+: :O2: :1600SW:
It bit me too. I haven't figured out the root cause of this, but my workaround was to copy the netscape start script to ~/bin and modify it to add:

Code: Select all

unset NPX_PLUGIN_PATH

before it runs the netscape binary. So, what is wrong with the plugin path? I am stumped, as the plugin path is definitely pointing to plugins under /var/netscape/communicator, and those appear to be the correct vintage.

I thought about digging into this, but really all I wanted was to get infosearch back in action, and try out wrp. wrp was a bust, but infosearch is happy, so I am not sure if it is worth digging more.
:1600SW: :O2: :O2+: :O2: :1600SW: