www.virtualacorn.co.uk/forum

Is there a known problem with adding the 256th file to a directory?
This is what happened on my system:
I was documenting a software library using !Stronghelp. I was using a
prefix to group the help pages on function name.
It happened that I had 256 (or more) functions with the same prefix.
When I tried to save the help page for the 256th function into the prefix
directory in the Stronghelp manual I got abort on data transfer messages
Internal error: abort on data transfer at &0391E544 from the filer.
I had to restart VA to get rid of the messages and access filer windows,
but I still couldnt access the affected directory from a filer window - VA
would just lock up, although !Stronghelp could access the files in the
directory and display them as Stronghelp pages, and I could remove the
directory from my system by dragging it into the ram disk and restarting VA.
To test it was not a Stronghelp problem I created an empty directory
(testdir) on HardDisc4 and ran this little BASIC program to create 255
small files in the directory.
root$="HostFS::HardDisc4.$.testdir."
FOR L%=1TO255
path$=root$+STR$(L%)
file%=OPENOUT(path$)
BPUT#file%,"This is a test file"
CLOSE#file%
NEXT L%
END
and then dragged another file into the directory and also got
the abort on data transfer.
I got the same result on my second VirtualRPC-ADJUSTSA system.
The same abort also occurred when I ran the program in VirtualRPC-Adjust.
The abort does not occur on my Risc PC running RO 4.02.
Has anyone else encountered this problem, and is there a way around it?

I've never had this reported before. There are limits on the number of files that RISC OS can display in a filer window (although I don't know what the limit is off hand) but it should be more than 256.
Can you check and see who/what owns that memory address?

Ive done some more investigating. I installed a fresh copy of VA which
works OK with the Harddic4 where the abort occurs. So, it may be a
configuration error. Just to double check, the new installation model.cfg file
has the following memory configuration
[Rom]
filename = "Romsets\RISCOS Adjust"
base = 0
range = 16Mb
[Ram]
base = 0x10000000
size = 32Mb
What is the maximum value that the [ram] size can be?

The maximum configured RAM is 256Mb. However its very unlikely to be memory related. Far more likely is that you have some patch module installed that's designed to alter the filer behaviour in some way - and it's crashing.
Using the copy that fails. Rename the !Boot sequence and then quit and er-load it (it will whinge, but from the green boot screen just go to the desktop). Now try your program. If it works you know its something in the Boot sequence.

Just to answer your earlier enquiry:
Doing a *modules reports
21 038FA200 02110654 InternationalKeyboard
22 039037CC 0411A014 WindowManager
23 03921AB8 00000000 Desktop
24 039227E0 022160C4 TaskManager
so I suppose that puts the abort address in the window manager.
I renamed !boot as you suggested, but I still got the data abort.
To investigate further, I have installed a fresh copy of VirtualRPC-AdjustSA
on another drive so that I have a clean boot sequence to use.
I ran my program, this time creating 256 files in the directory. I did not
get the data abort.
However, since I have the [Ram] size configured to 256Mb in VA I decided to
try different [Ram] sizes with this clean boot sequence.
There was no data abort with sizes of 16, 32, 64 and 128Mb, but with the
[Ram] size set to 256Mb the data abort occurs. It is exhibited by clicking
on the directory in the filer window.
I tried renaming !boot on this fresh copy and still got the data abort.
If I reduce the configured [Ram] size to 128Mb the data abort does not occur.

I've also installed a clean copy of VA on a second machine,
configured [Ram] size to 256Mb and it has the same problem.
Once the data abort occurs the filer fails to function correctly.
The data abort also occurs when I try to access any directory
containing 256 or more objects for the first time from a filer window.
For example, using !DrivePop to open a filer window for the main C drive
and clicking on the windows installation directory.
One more thing I did was to add enough files to HostFS::HardDisc4.$
so that it had 256 objects in it. After that I was unable to open
HostFS::HardDisc4.$ because of the data abort.
Both machines I have tried this on are otherwise stable but the problem is
consistently produced just by trying to open a directory of 256 or more
objects in a filer window.
Since both my systems have the problem, it looks more like a general
problem affecting the filer and VA with configured [Ram] size 256Mb.

Given that the only alteration is the amount of configured memory I am inclined to suspect that the problem isn't HostFS, or VirtualAcorn, but RISC OS itself.
I would suggest asking RISCOS Ltd if they have had any similar reports from other users (of either virtual or real machines).
It might be that some cache somewhere is rolling over and needs increasing, but I don't know which/where.

I seem to recall having a similar problem with some version of select on my RiscPC with remote directories of over 256 files on my linux box. This occurred when I installed more memory, IIRC over 64 or 128 meg. I can't remember if this was fixed in later versions of select.
So it's possibly a Select problem which was reported by several users in the select mailing list.