Fatdog64 710 Beta2 [CLOSED]

[spandey]

My problem is Ram layer is not worrking with savefile. Here's my situation
1. I have main files on sdb2 which is SSD.
2. SAvefile is on sdb1 which is connected to USB2
3. I boot from SSD using savefile=ram:uuid:59d7a756-c1c6-4f07-a1ee-8ca361dfe507:fd64save waitdev=5

The problem is the savefile is loaded but RAM layer is missing. How should I investigate this problem? Any clues?
[/quote]

[jamesbond]

spandey, please run Bug Report from Control Panel -> Utilities. Upload the resulting tarball somewhere. If you don't want others to see it, PM me where I can get that tarball.

cheers!

[prehistoric]

@jamesbond

I'm now pretty thoroughly convinced the problem I reported is confined to one machine and is a symptom of a hardware bug. It applies to anything that pushes block IO close to maximum rates, and is not always in aufs. It remains difficult to reproduce.

The latest change took place when I was living with windows open and the temperature dropped. (It's approaching winter up here.) After an especially cool night the problem went away temporarily. Cooling electronics to operate at higher speeds is a standard practice.

Afraid I still don't have those tests with different kernels. My fight this weekend has been to recover networking after a change of ownership of my ISP.

Anyone want a book titled "When Bad Networks Happen To Good People"?

[prehistoric]

Very minor bug in Fatdog 710b2: color of battery icon does not always change properly from discharging to charging. While working on my networking problem, I left a laptop running 710b2 connected and running overnight. If it had not been charging, it would have drained the battery flat in hours. Battery indicator still shows red with 99% charge.

Good news: I ran Fatdog 710b2 on yet another machine (HP Pavilion a1520n) given to me to remove data before recycling. It worked fine, and I was able to use rsync to copy all the jpeg files to a flash drive without even looking at them, as recommended by MochiMoppel. Handed this back to previous owner, rather than anyone else. (That way I can't be accused of distributing porn.) I intend to put Fatdog 710 final on this machine before giving it away.

[jamesbond]

@prehistoric, thanks for the tests.

Your story reminds me of my old computer. It was a PC XT clone with dual floppy drive. It worked very well generally, except that it had one very annoying defect - when writing to the disk, once a while it would replace one character with a binary zero (NUL). This didn't happen very often, but it did happen. The result, of course, file corruption; executables crashed when run, data files sometimes show artefacts or even couldn't be opened if the corruption hit the critical file zone. It was rare, but the fact that it happened made me always worried that any data I saved today cannot be opened tomorrow. This was during the days where tools like md5sum didn't exist (on that platform); so I couldn't even detect which files got corrupted, and which weren't, so it made it so unreliable. I had to live with that for 3 years before finally migrating to i386 with worked flawlessly.

RE: battery meter icon:
a) if it is full (100% charge) - green
b) if it is being charged - blue
c) anything else - red.

cheers!

[mories]

RE: battery meter icon:
a) if it is full (100% charge) - green
b) if it is being charged - blue
c) anything else - red.

Vattery seems to have a problem when using locales other than C or en_XX. If you run with LANG other than C or en_XX colors are not displayed.

[jamesbond]

@mories - I have an alternative battery applet. It is called fatdog-battery-applet.sh. It's less configurable than vattery (you need to edit the script do carry out the actions etc) but may work better in your case.

[prehistoric]

...
RE: battery meter icon:
a) if it is full (100% charge) - green
b) if it is being charged - blue
c) anything else - red.

cheers!

The problem was that it was being charged, or at least holding its own, and showed red. I left the machine connected to the mains for several days, and that's the way it stayed. Probably has an old battery, but we need some way to tell if it is getting power. There was no visible change when I unplugged it, unless I remembered how to read the several LEDs on the laptop. Way too easy to have a loose connector that causes a full discharge and crash.
-----

If we are going to trade digital war stories about intermittents, I'll go back to a problem with a hard disk with a large removable cartridge (14"). It stored a whopping 10 MB, a big improvement on the previous model. This didn't seem like much until that marginal problem wrecked file structure, and I had to go through every disk block with a hexadecimal editor to piece the system back together. Never again!

[dr. Dan]

Hello All.

I read a history of Fatdog64, and I say thanks again to the developers!

When I first installed Fatdog64 700, I think I remember it had a battery meter which gave me an estimated time left. After my explorations with other distributions, when I reinstalled Fatdog64 from a freshly downloaded image, it was no longer so. Or, my memory is off. The learning curve is mighty steep at the start. Either way, can a script be modified to include that information? I might be able to figure something out, in time, but I'm not looking to be a developer right now, and my skills are limited. And for that matter, a migration script (from 700/702) would be of greater use.

Your time is appreciated.

Dan

[jamesbond]

@prehistoric

The signals all come from the laptop.
I have a laptop that has the option of stopping charging when battery capacity reaches 80%.
In my laptop, this is how it behaves:
a) battery discharge: red
b) power plugged in, battery charging, capacity less than 80%: blue
c) capacity reaches 80%: red --> this is because the battery is no longer being charged, and it isn't 100%.

If I remove the restriction and let the battery be charged up to 100%, the icon will be blue up to 99% and turns green once it has reached 100%. When unplug the power, it stays green as long as it is 100%, and as soon as it drops to 99%, colour changes to red.

Your case sounds very similar to me except that in your case, it is the battery controller that probably decides to stop charging when the battery cap reaches certain limit.

Now your concern is real, but the battery indicator just shows the state of the battery (full, charge/discharge), it doesn't actually tell us about the state of the power. Just because power cable is plugged doesn't mean the battery is being charged.

We need another indicator to state that power is plugged in, regardless of the battery state.

The state of power (in my laptop, at least) can be found here: /sys/class/power_supply/ACAD/online --- if it is "1", then power is plugged, if it's "0" then is not plugged. The same information may be available here as well: /proc/acpi/ac_adapter/ACAD/state

Any takers? One can use "sit" to display icon in the systray. Someone could start from fatdog-battery-applet.sh script and work their way from there.

-----

Your story about patching up disk blocks using hex editor is horrible. With 10MB disk I could imagine you spent hours doing that. My silly floppy drive also write NULs to FAT, and of course it resulted in truncated and/or invalid files altogether; but 360K two orders of magnitude less than your problem ... also, I did the FAT patching using a tool, rather than plain hex editor (viewing 12-bit FAT using hexeditor is really no fun); so I suppose that makes your problem at least 3 orders of magnitude more difficult.

@Dr. Dan:

We used a few battery state display on 700. The earliest one was batterymetter, later we changed that to qbat, and on the latest release we used vattery (which is also used in 710). The one that would display the remaining time is probably batterymeter. I can't remember off-hand why we abandoned it; but I remember that we dropped qbat because it has the habit of suddenly hogging 100% of CPU time for no good reason.

Cam a script be created to display remaining time? Yes, possible, but since the information doesn't seem to be provided by the system anymore; doing this (properly) is quite complex. one needs to model the battery discharge voltages against expected uptime, also based on the time the system has been up, etc.

If anybody wants to take the chanllenge, welcome.

I will talk about migration later. Bed time now.

cheers!

[prehistoric]

@prehistoric
...Your story about patching up disk blocks using hex editor is horrible. With 10MB disk I could imagine you spent hours doing that. My silly floppy drive also write NULs to FAT, and of course it resulted in truncated and/or invalid files altogether; but 360K two orders of magnitude less than your problem ... also, I did the FAT patching using a tool, rather than plain hex editor (viewing 12-bit FAT using hexeditor is really no fun); so I suppose that makes your problem at least 3 orders of magnitude more difficult.

This post is off-topic, strictly speaking, so most of you can disregard it.

This experience took place before Microsoft existed. That system had a File Address Table, which was still a generic term at the time. This was corrupted, as was a bit map of used/free blocks on the disk. The design of the file system had been stretched from one not too different from that used on a floppy disk, but people weren't yet thinking in terms of superblocks and other kinds of redundancy. We were supposed to keep full backups on mag tape in case of catastrophe. The problem developed at the end of a working day, but before backup.

It's hard to believe today, but many people were doing software development on that shared minicomputer. Rolling back to a previous backup would have cost us days of work, and we were near deadlines. A major problem was the sheer panic which set in when we knew the disk had been clobbered.

One ray of hope was that the files we really needed were source code files. We could restore binaries from the minicomputer vendor without problems. If we had source for our custom programs, and the proprietary compilers, we could recreate all the binaries in a few hours of operator's time. My main problem was deciding if a block with text from our proprietary source language was in the file system or the free list.

Once the processor was repaired, and I had pieced the system back together, it was just a matter of a late shift for operators to rebuild those binaries we needed. I even got a little sleep that night, which was better than some company crises.

This was typical of the exciting world of high-tech start-ups. Despite vast changes in technology, the sociology of start-ups remains similar.

[TeX Dog]

Yeah, you had it easy, I had to write octal code from a bank of actual physical CORE memory because someone NOT ME did not load the longest blank tape into the teletype

[Sage]

I may still have a bag of wound ferrite memory cores somewhere in the junk box? Also saved some 4-transistor flip-flops used in the first Ferranti 1904 all-solid-state mainframe ca.1962. Thought of setting them in acrylic and selling as trinkets. Goodness knows what other memorabilia lurks in my caverns. However, unlike the SW gurus on this Forum, I was a mere user back then and an avid HW constructor (one of my patents influenced the electronics construction industry).

[jamesbond]

@prehistoric, @TexDog: Thanks for the interesting stories.
@Sage: Now I understand your pre-occupation with Raspeberry Pi!

[prehistoric]

Here is where my exposure to core memory originated, though at the time, the machine was not in a museum. The author has taken the liberty of pretending there were only 8 planes of memory, ignoring word marks, which would take another page of explanation. This was in no way typical of modern computers. It took 11.4 microseconds to read one character of memory, and with the exception of instructions like stop, written as a period, most required a number of characters.

Oh, BTW, those index registers he talks about were not in the earliest models, so instruction modification was de rigueur.

@Sage, do you have any germanium diodes removed from early digital electronics when these were replaced with silicon? Since the originals cost $1 apiece, a friend of mine hung onto these. I'll bet there are more still in his house, (if anyone can find them. I would not care to place bets on what else will be found when the place is cleaned out. ) I gave some to a ham friend who built a crystal radio with one.

Now back to your regular discussion.

[Sage]

do you have any germanium diodes

Funny you should ask. I thought I'd given them all away when I was doing my history of the electronics industry talks to school kids. But, no! I've still got a few boxes containing, eg 2G339, Y401J1, XA701,, others, some slices of single crystal (used in my lectures on chemistry of prep & pulling), the flip-flops seem to use M1343 which may be silicon?, 4only SQ military quality thermionic valves (bought on surplus market), some other stuff in a different location. Not sure about those ferrite cores, though. Anyone wishing a specimen could PM me. Ge slice is NOT for sale, though, and only have one left.

Found a couple of those infamous 'Red Spot' trannies; they're in long thin gold cases, which used to mean they were 'high frequency' (RF that is!) £2 each at the time. Could be worth ten times as much now as most folks managed to blow them up. Yes, the ferrite cores were left at my last abode, ie gone.

[TeX Dog]

Hay Sage (sorry JB & kirk) I once worked for a company here with a parent corp in GB, they got there start making electrical fires ( wall switches) HOW did that term ever come about? It was odd we had to practice walking the halls GB style so the CEO during visits would not be making it too obvious he was passing WRONGLY sided.
They did tell us to drop the fake English accents and stop drinking our soda-pop cans with our pinkies extended while training to walk like a Briton.

[chapchap70]

I got a new laptop to play with and tried this. (Actually, it's old; someone was going to dispose of it... Quad 4s are old rubbish these days here.)

Seamonkey and LibreOffice take 5 to 6 seconds to open on this machine.

I compiled the latest Hiawatha Webserver in 702 but the package works for me with this. You can find it here... http://murga-linux.com/puppy/viewtopic. ... &start=133

[TeX Dog]

I got a new laptop to play with and tried this. (Actually, it's old; someone was going to dispose of it... Quad 4s are old rubbish these days here.)

Seamonkey and LibreOffice take 5 to 6 seconds to open on this machine.

I compiled the latest Hiawatha Webserver in 702 but the package works for me with this. You can find it here... http://murga-linux.com/puppy/viewtopic. ... &start=133

Is that fast for you? Do you want it near instant? there are not much talked about boot flags but if you have over 2G RAM there are faster options built in. How ever coming from windows or other bloated linuxes that does seem fast to some, its slow to me

[chapchap70]

Is that fast for you? Do you want it near instant? there are not much talked about boot flags but if you have over 2G RAM there are faster options built in. How ever coming from windows or other bloated linuxes that does seem fast to some, its slow to me

Actually, the 5 to 6 second time was the first time opening the apps. LibreOffice opens for me in 2 to 3 seconds and SeaMonkey is a little slower.