Citylink - useless, useless, useless.

[Windmill]

Someone described, 40 or 50 years ago, how to insert invisible self-propagating code into compilers and login programs.

Last year my cheap Emachines PC became unusable because its VGA output became scrambled (and I didn't have an HDMI monitor to check the HDMI output). Booting from a Knoppix CD didn't help; the display was still garbled.

I decided that, Real Soon Now, I'd make a bootable memory stick which I would set up so as to allow me to hook up a serial link between the Emachine and my cheap Gateway laptop. Then I could investigate what was going on inside the Emachine.

Months later, the laptop suffered a similar fate; scrambled output on the screen. Again, booting from the Knoppix CD, which worked elsewhere, didn't help. But its external VGA output still worked, so using an external monitor I (finally) created that bootable memory stick for use on the Emachine.

Then, sitting the laptop, external monitor, and Emachine side by side, I booted the Emachine (typing blind) then connected the serial outputs of the two machines (actually USB-to-serial converters) together using a null modem cable.

The instant I connected the cable between the two machines, before I had time to attempt a login from the Gateway to the Emachine, *both* machines started to display normally.

Far more than bizarre.

Both use an NVidia chip; I wonder what else is in them besides VGA circuitry?

I've seen some strange things in the past (such as a PDP10, used by AECL for reactor calculations, whose FDIVR instruction occasionally produced random results but only at night - it turned out that it failed over only a very narrow range of temperatures) but I can't quite convince myself that this VGA problem is hardware-related.

The story goes that NVidia had a temperature problem with some of their chips, but I'm having trouble believing that is the whole story.

[Windmill]

Some people point out that, despite an increase in the number of people 'in work', national productivity is down.

Large organisations seem keen to make individuals do more, (for example on-line) but the result might be a net loss of productivity.

[Man at B&Q]

My 14yo found my Java book on the bookshelf and figured out how to get going without any help at all.

MBQ

[Man at B&Q]

Or have a soft core processor in an FPGA.

MBQ

[Man at B&Q]

What were they on/thinking?

MBQ

[Man at B&Q]

What sort of different? Which manufacturer?

You cannot download the configuration bitstream to the wrong device, let alone get as far as running code on a soft core.

Are you sure it wasn't just a different speed grade of the same FPGA?

MBQ

[Roland Perry]

In message , at 09:19:45 on Fri, 25 Jan 2013, Nightjar remarked:

One of the "blue chip" couriers I used in the 80's managed to card a 5* hotel with 24x7 concierge and front desk.

[Man at B&Q]

Well, everyone has to take a dump sometime or other.

MBQ

[Roland Perry]

In message , at

02:59:11 on Sat, 26 Jan 2013, Man at B&Q remarked:

If you are suggesting the hotel was unmanned as a result, then you'd probably be wrong. It was large (over 1000 rooms) with people around

24x7. Think of somewhere like the Hilton on Park Lane (although this hotel was in the USA).

[Man at B&Q]

I didn't really expect anyone to take it seriously.

MBQ

[Roland Perry]

In message , at 10:45:41 on Fri, 25 Jan

2013, "Dave Plowman (News)" remarked:

The solution here is to have three of the wireless doorbell "receivers", one of them in a 13A socket right next to the front door.

It's *much* louder than the traditional wired doorbell that came with the house.

[Dave Plowman (News)]

No it's not.

[Roland Perry]

In message , at 12:43:58 on Sat, 26 Jan

2013, "Dave Plowman (News)" remarked:

I must have been out when you came round and listened to them both. But as I disconnected the wired doorbell the same day I fitted the wireless one, it must have been very well timed by you (to do a direct A-B comparison).

[Dave Plowman (News)]

Sorry, miss-read your earlier post. However, if you already had a wired door bell which worked but wasn't loud enough, why not just change it for one which was? And or add another? Got to be less hassle than adding three wireless ones.

[Owain]

My Bedlam telephone warbler says I must wear hearing protection anywhere within 3 metres of it.

Owain

[dennis]

It was some search that I never really understood why it was in microcode as it wasn't really possible to allow such a long instruction to run and it wasn't pre-emptible IIRC.

None of the actual real programs used it.

I suppose someone thought it was a good idea when the instruction set was designed.

It was before my time, I just had to design some of the hardware to run it. It pre-dated microprocessors BTW.

[John Rumm]

The processor in question was in an AVR device - but I don't think that is in the FPGA they were having issues with. It may have been a Xylinx... I can find out tomorrow if you really want to know?

This was a compilation problem - i.e. re-targeting existing code at a new device. So the binary was fully compatible, and in theory it should have worked. But you know "in theory" does not always hack it.

(I tend to find flaws of that sort are not uncommon in the embedded development environment. Pretty much every tool chain I have used has had at least a few "special" features that one needs to learn to cope with. Most of the cross compilers I have used will generate bad code in certain circumstances... its probably just reflection of the size of the user base).

To be fair I was not the one doing the debug in this case and am only reporting events second hand... I suspect it was a similar but larger FPGA. Since the job was timing critical I doubt they would have gone for a different speed device.

[Man at B&Q]

More a general interest. My real knowledge is Altera.

I think something is being lost in translation. Are you talking about re-targettng the FPGA bitstream? There is no "binary compatibility" between different devices. Compatibility is at the RTL (VHDL in this case) level.

This sounds more like a processor toolchain but you soid the processor was separate device?

MBQ

[Bob Eager]

The stuff I used was loaded in during machine initialisation. 60 bits wide. And I didn't have an assembler!

[Bob Eager]

I remember that (at the time). Some details here: