OS4.1 bugs

When I have swap enabled and trying to open a 700MB file with datatypes system makes whole system to freeze. I left system running two hours to make sure it wasn't just slow down. (512MB real RAM.) If I disable swap then there's not problems.

Edit: At first it looked like it was Sprite datatype causing this problem. But testing continues...

There's also another thing to just mention. If I open an archive file (7z) containing 700MB file it looks like system was frozen first but it's actually only slow down and system works fine after a couple of minutes.

@TSK

Its tricky to know if the system really has freeze or if its just for a time. Sometime only the mousepointer freezes and the system goes on. I once tested to move the mouse from ps2 port to usb while the freeze is and then i could move the mouse again. If i then put it back on ps2 port then i cant move it. Strange one :D

@Kicko

You shouldn't connect or disconnect ps2 devices when you have power on. You might break your mobo.

(Edit: I edited my previous post.)

@TSK

Quote:

Might be that the DataType in question doesn't actually support swap mem, and also has a bug where it doesn't check memory allocations properly. Which DataType were you using?

Quote:

I think the swap mem runs at a very high priority (presumably Exec's priority). The slowdown only occurs at the start of the process - which will be when the data is being loaded into swap mem. Once it is trying to parse the data (read it from swap) everything is responsive again.

You can tell the difference between slowdown and a freeze by checking if there is any disk activity. If there is, the swap partition is probably being utilised and it is worth waiting. If not, there's a 99% chance the whole thing has locked up. You can use DiskLED.docky if you have no real disk activity light wired up (eg if you have a SiI3112 like me which doesn't support it)

Chris

@TSK

Hey, if OS4.1 has bugs in it, couldn't you just get rid of them with some insect spray or something?

Nathan

@Chris

I forgot to say it previously but I made several tests to find that bug. If I try to open that file as a picture then system crashes but if I try to open it as anim or movie then there's not any crash. (I haven't tested sound DT.) It doesn't matter if I use an Intuition or Datatypes function. (The file itself is actually a CD disk image.)

Quote:
Yes, it looks so.

Quote:
Yes, that's how I realized it. But we are so used to how responsive Amiga usually is so it's easy to start thinking first it has frozen when nothing moves anymore on screen.

@amigamaniac_aus
Quote:
Heh heh ! Btw. Do you know why people started to use word "bug" as a software bug in the first place ?

@TSK

I don't understand how you have managed to try to open it as different types if it is a ISO image, it shouldn't open at all unless you have a binary.datatype.

If I take what you say at face value then it is picture.datatype that has a problem, but you must be using some subclass to get it to load. I don't really understand what you are doing - have you tried a real 700MB picture in a picture datatype supported format?

@TSK

He haw! There's always one, and it's usually me Still, I had a laugh everytime I checked the post for a reply, so at least it entertained me

I did some research on where the term came from. There are several sources. I couldn't really identify the first people/person to use it. Some info about software bugs on Wikipedia - the world's most reliable source of information Worth a read.

Nathan

@TSK

He haw! There's always one, and it's usually me Still, I had a laugh everytime I checked the post for a reply, so at least it entertained me

I did some research on where the term came from. There are several sources. I couldn't really identify the first people/person to use it. Some info about software bugs on Wikipedia - the world's most reliable source of information Worth a read.

Nathan

@amigamaniac_aus
I checked Wikipedia too. It looks like that when they found a moth from a computer relay it wasn't the first time when the word "bug" was used. But those guys had a good sense of humour. There's even a picture of that log book where they taped that moth. It's funny: "First actual case of bug being found".

@Chris
Alright. You asked. So I answer long version this time.

Quote:
Of course it doesn't open because it's wrong file type. But datatypes library shouldn't cause a crash when examining a file.

Quote:
I don't think I have any such big picture to test.

Quote:
You can force datatypes library to try to open only certain types of files using DTA_GroupID tag. For example this is much faster...:
if (DTFile=IDataTypes->NewDTObject(filename,DTA_GroupID,GID_PICTURE,...,TAG_DONE))
else
if (DTFile=IDataTypes->NewDTObject(filename,DTA_GroupID,GID_TEXT,...,TAG_DONE))
else
if (DTFile=IDataTypes->NewDTObject(filename,DTA_GroupID,GID_ANIMATION,...,TAG_DONE))
else
if (DTFile=IDataTypes->NewDTObject(filename,DTA_GroupID,GID_MOVIE,...,TAG_DONE))

... than this:
DTFile=IDataTypes->NewDTObject(filename,...,TAG_DONE);

The latter is significantly slower. (I guess in the latter case the library is trying/testing every single datatype you have installed.)

My program is trying to use scaled thumbnails replacing icons. So originally my program loaded icons first so a user can browse a drawer when a computer is trying to detect all file types in the background. I had to spend few hours to make many changes to my program so it will try to detect all file types first before loading icons (and using default icons in the meanwhile) to work around that problem.

My guess is that picture class is trying to load the whole file into memory and with paging enabled Exec will give such big block of memory to the class. But anim class is trying to load only one anim frame at the time so it doesn't try to allocate such huge amount of memory. But I've never seen the source code of the datatypes library so this is only a guess.

@Chris + all
I was going to start a thread something like: "Did you know AOS4.x can do this" type of thread some time ago. But I didn't do it. So if you don't have WBPreview installed already download and install it from os4depot. Then open some drawer containing for example mp3 files which contains embedded cover picture on Workbench. Select any file in the drawer. Open a preview. If any of your files don't have such embedded pictures then download this NewsTek podcast and try again and be amazed (Hint: "Cover flow" on Amiga.)

So it's a very good idea try to open f.ex. audio files as picture files (I must say I found that "feature" accidently.) One of my friends sent a Powerpoint file to me in email some time ago. It had such embedded "cover picture" too. (Those two files are the only non-picture files on my hard drive which have a picture embedded.)

@Hans
This comment is from os4depot: Quote:
I don't know still how to make datatypes objects to work inside Reaction layout (or if it's possible at all).

(Edited)

@TSK

Quote:

If rtg.library would have a "nasty bug" this would have been spotted long before the actual release of OS4.1. Noone else is spotting this, so its very likely a problem only on your side.

You should check the usual suspects in case of random crashes:

- If on A1, check your battery. The battery *has* to deliver 3V. Anything lower than that can lead to unpredictable problems. Of course, the lower the current the more problematic it is.

- Check your memory. More than often faulty memory caused random crashes for many people. There is a memtester on os4depot which you should run after booting without startup-sequence for a long time (e.g. over night). A quick test for heavily broken memory would be to just fill up the RAM disk. E.g. copy a 100MB file into RAM Disk over and over again until its filled up or crashed. But as said, this only happens for heavy memory errors.

- Last not least: Make sure your graphics cards cooling works and won't let become the gfx chip too hot. OS4.1 with compositing makes much more use of the gfx card than older versions, so with an improper cooling your gfx card may easily overheat..

edit: double post

@Cyborg

*check your battery: It's not long ago when I changed it and the previous one worked many years.

*Check your memory: I've done that already many years ago.

*graphics cards cooling: It has passive cooling.

@TSK

Check your Datatype descriptors. datatypes.library won't open the entire file before checking it - only enough to identify the file. Probably you have a buggy descriptor which is reading the entire file.

Disable them all and add them back one-by-one until you find the culprit.

@TSK

Quote:

That doesn't mean it can't be bad again. My first A1 battery also lasted almost 4 years, but the second was dead after 2 months again, the third is now going for more than half a year...

Just check the current of the battery to be sure. It isn't much work but it rules out one possible problem.

Quote:

Oh and because you checked your car already many years ago it can't be broken *today* or what?

Come on.. you have the problems *now*, so you have to make the checks *now*.

Quote:

That doesn't mean that it can't get hot! Man, exactly the opposite: without a fan it can get even hotter... maybe you should place a fan over it just to be sure.


.. but .. wait.. i tried to help you with an OS4.1 crash and told you 3 common problems which may perfectly be the reason for the crash AND i told you how these possible reasons could be ruled out.. but you don't seem to actually want help because things like "i checked my memory years ago and it was fine" or "my 1st battery lasted for years, so the new one can't be worse" or "nah, my graphics card is passively cooled, so it simply can not get hot" is just .. well.... utterly bullshit!

sorry to say so, but this are the facts: if you ask for help, accept the offered help and do your parts on finding the reason. If you aren't willing to do a simple current-check of your battery, then stop complaining about problems or crashs right now! Simple as that..

P.S: Yes, i am pissed..

@TSK

Could you post here the output of both of these commands?

dir SYS:Devs/DataTypes/~(#?.info) MAXCOLUMNS=1
dir SYS:Classes/DataTypes/#?.datatype MAXCOLUMNS=1

Most likely the problem is with one of the descriptors or datatype subclasses.

@Chris

Quote:
I found the datatype to cause this trouble. It's Sprite datatype ! I'm sorry I put this in wrong thread.

(Edited)

@TSK

Doesn't surprise me - that descriptor isn't great because it isn't easy to identify RISC OS Sprites, so it does some basic byte checks and that's all. I'll see if I can make it a bit more robust - and stop it crashing if it does misdetect a file. In the meantime, can you send me the first 100 bytes or so of that file? Load it into FileX and screenshot it if that's easier.

Also give the attached version a try.. I doubt it will make a difference but it might.

Thanks

Attach file:

---

Link only for registered users

@Chris

Edit: Forget this post and jump directly to the next one !




I tested your new version but it didn't help. I don't need actually this Sprite datatype. I have a habit to download and test everything from os4depot and if it doesn't give any trouble I just leave it there installed. But I don't actually need it.

The file I'm using to trigger this problem I think I used MakeCD to create it and it contains audio tracks only (audio CD). You can make a test program for yourself. Use either if these two functions:
IIntuition->ObtainBitMapSource(filename,TAG_DONE);

IDataTypes->NewDTObject(filename,DTA_GroupID,GID_PICTURE,PDTA_DestMode,PMODE_V43,TAG_DONE);

When I was trying to find the problem the first time I put a lot of printf's into my program so the computer didn't ever reach the next line after either of those two function calls. I tested them both separately, of course.

Update: I tried to open that file with FileX (2.5 for AOS4.x) with the same results. I waited 4 minutes before resetting. Then I decided to test another image file just in case my testing file was somehow corrupted. I tried to open an .iso file, which I downloaded from AROS website, with FileX and all dockies froze but mouse pointer was movable. I waited five minutes. Then I clicked anything on screen with left mouse button and mouse pointer froze too. There wasn't anything in debug buffer after rebooting.

Testing my hardware will take several days because at the moment I don't have too much time for my computer. And I want to test one another thing first.

@Chris

Forget my previous post.

More update:
APTR memptr=0;

memptr=(APTR)IExec->AllocVec(700000000,MEMF_SHARED);
if (memptr!=0)
{
printf("Got some mem: %ld\n",(uint32)memptr);
IExec->FreeVec(memptr);
} /* if */
else
printf("Mem alloc failed: %ld\n",(uint32)memptr);

If I'll copy 700MB file to RAM disk and delete it there's not any problems, if I don't try to start any app right after.

I can test my hardware anyway when I have enough time to do so. (If I can remember I had trouble to get my computer to boot again last time I was messing with the battery. I have to backup a lot of important things first !)

Tested:
I tested booting a clean OS install and lowering my hard drive to PIO mode. No difference !