The Eagle Phantom Connection Menace

The metric switch is working fine (its only the page size) but it easely triggers a user error. I found the cause of the problem quickly but my boss had been pulling his hairs out for years.

That sounds more like bad PC hardware or time for re-installing Windows. When I need to use customer specific software I use a different PC (A-brand) with a freshly installed Windows.

How could that be if _all_ other programs (and there's a lot) work fine? Later I heard that several other people have experienced some of the same errors. No, I absolutely do not believe it's this PC. Neither did Cadence support.

Besides, Eagle works crashlessly for several years. So they must have been doing some things better.

When I was still in the PC repair business I once had a client with a bad PC. He took it back to the shop 4 times but every time he got it back with 'problem not found'. Finally he had me diagnose it. I found out that one program in particular didn't work which did work on other PCs. I wrote a report on how to reproduce the error and he got a new PC. Some people say a PC is a PC is a PC is a PC.... I know from experience they are terribly wrong.

Errors can be very elusive. Last year I bought a new PC. In some rare cases it failed to compile software. After long testing with memtest86 it turned out to be a bad memory module which failed in a very particular bit pattern. So even when everything may seem fine there still can be a problem in the hardware.

Ok, yes, that can happen. But I have some software here that completely maxes out all memory this PC has, and then some. Never a crash, it just gets slowed down when the paging starts.

Also, I have managed to crash Orcad many times at clients. Lots of other PCs. Nobody was particularly surprised when it happened.

Of course the immediate poke is to ask whether you have discussed this = the vendor? Can you replicate the phenomena? Reliably?

?-)

rather

not

only

reached

Kind of makes me wonder if some portion of those crashed were windows and DLL hell rather than orcad itself so much. I think i will have to hunt down an old version to run in a dos box

?-)

Even if it was related to Windows and DLLs, a company is supposed to test their products in the real world.

Nup.

Can you replicate the phenomena?

Nup.

Reliably?

Nup.

The error happened on a board that I had been playing with very extensively. Ripup and route, edit schematic, save, change stuff, etc, etc.

I'd be very hard pressed to duplicate all the steps. :)

Just now, I tried to create a version of one of Eagle's demo boards to show the problem. The closest I got to the issue was when I placed a wire on the schematic, routed it on the board, saved, ripped and routed the same trace using a few curves, saved then deleted the connection in the source file. Then I looked at the board and found that most of the hand-routed trace remained; there was no connectivity because one section of the trace disappeared. All was better after I ripped up the errant copper on the board.

Some miracles are not perfect. :)

--Winston

and

hunt

Not since MSWindows hit the street.

?-)

There are a lot of companies that are able to do it. Cadsoft being one of them. Or take a look at LTSpice which is rather demanding on the OS and the PC. I have yet to see it crash and I have ridden it hard.

Somehow there isn't much in the middle, there's lot of software that never crashes and then there is other software that crashes all the time. So obviously one group of companies is doing something right that the other group isn't.