Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - Valery Sedletski

Pages: 1 2 3 [4] 5 6 ... 24
2Martin: Yes, brackets in pathnames may be illegal symbols

> I was wondering also a different thing. You call your file with version 5.0.51, but currently the official vbox OS/2 guest drivers are on version 6.1. (which I guess there is not much difference). Are you using a current copy of the addons source code? Do you have it sync with the official repository?

No, My version is different and not related to the official version. They have completely different source code and can't be merged. Of course, I didn't
backported anything from 6.1 additions to 5.0 ones.

> 3) Copying files from the WPS (from the shared folder - Windows Host, to the WPS desktop) gives me some problems too.
Some files give a path can not be found error, other just lock the WPS on the copying dialog box.

Maybe, too long filenames or illegal filenames? Did you checked that?

> 4) Maybe it is related to problem #3. I can open directly from the host some files like PNG, PDFs, and some wpi installers. But EXE WarpIn installer give me this error when I run them from the shared host.

Maybe, some access right-related problems? I had some problems trying to rename a file but it appeared that these files had access right problems. I had a JFS partition
under Linux, and a file had a "readonly" attribute, so I was unable to rename it under linux. So, it was a problem outside VBox. Maybe, this file has access problems on
the host? Can you check a different guest OS with the same shared folder? Maybe, these problems reproduce there? Also, did you checked if the .exe can be
started from FC/2? Is there any difference from the WPS window?

PS: Did you tried to look what is written in the COM port log?

PPS: Tried to copy files with the WPS. There are some problems with that. If I drag a file with WPS from a shared folder to a ramdisk, I got an "SYS0225: The extended attribute list size is incorrect". After the second attempt to do so I got a kernel trap. Looking at the COM port log, I see that there is something wrong
with FS32_FILEINFO. It looks like I repaired FS32_FINDFIRST/NEXT onlym but not FS32_FILEINFO/PATHINFO. Need to check this further.

Hi ppl!

Today I updated my version of VBox additions to the following version:

Now it has a long-awaited fix present: an empty EA's generation fix. Now dummy EA's generated correctly.
This makes the WPS Drives object to work as it should. Previously it began populating the folder window,
then in the middle of process, it wrote an error about "No objects that are meet the specified find criteria...".
Then you pressed the OK or Cancel button and the window closed. Now this error is fixed.
Please test.

2Andi B.:
> LAPIC ID's are consecutive - this is an old Intel dual-core machine which is still my main system although it should have been retired since years.
If so, then it may be not from {0, 1} set. VBox assumes that CPU ID's should be from a 0, ... Max. CPU ID set. If this is not true, this error could occur.
If you don't want to touch this machine, then you could leave everything as is, and use VBox r143 from your another machine. It should use a single
CPU only because of no SMP support in VBox.

Indeed, testing OS/4 kernel is easy thing. I cannot expect you can break much. And you can easily switch kernels with QSINIT or OS/4 loader. You
only need to replace the old IBM loader with new one, and set up your os2ldr.ini. You can boot the old kernel and newer kernels from the same menu.
And QSINIT is backward compatible with older kernels. It should even be able to load Warp3 kernels successfully (it was tested by its author). So,
there should be no problems here. At the extreme case, you can chainload on older os2ldr version with 'restart' directive.

> Anyway I think currently the biggest problem is the amount of lower shared memory VBox needs. I think we should find ways to reduce that at first.
The 1st thing you need is to set up correct values of VAL, THREADS and PROCESSES in your config.sys. Plus see if you did not used too much
JFS cache. For example, I have

IFS=D:\OS2\JFS.IFS /LW:5,20,4 /AUTOCHECK:* /CACHE:128000

Here, 511 threads and 128 processes should be sufficient for most cases. The default values from config.sys are too big and allocate
about 200 MB for Thread control blocks and PTDA's, which may eat up much kernel memory. Also, here you can see 3 GB of VAL
and 128 MB of JFS cache. You need to check (with theseus) that there are some 100-200 MB of kernel memory left after these settings.

Plus, you can try loading most VBox DLL's to high memory. I prepared a highmem.cmd script for that in the VBox distribution for that.

2AndiB: The -1018 error means VERR_INVALID_CPU_ID. You're not the first person which encounter this error. You're probably using an IBM's kernel.
IBM's kernels don't have a means to determine a CPU ID, which is calculated as an LAPIC ID by default by VBox. But VBox assumes CPU ID's to be
consecutive, which is not always true in case of LAPIC ID's. So, if they're not consecutive in your system, this error can occur. On some systems, like
mine, LAPIC ID's are consecutive (there is no hole). So, this error doesn't occur in my system. So, I'm lucky. Unfortunately, there is no reliable way
to determine a CPU ID on an IBM kernel. OS/4 kernel contains a KEE function to get a current CPU ID, so no such problem with it. So, I can advice
you using an OS/4 kernel here.

BTW, VT-x (a hardware virtualization) support is working with OS/4 kernel, only. So, this is another bonus of OS/4 kernel. OS/4 kernel is required
because of a "CPU rendezvous" KEE, needed to run code on all CPU's in parallel. IBM's kernel also lacks this feature,

PS: Version 5.0.6_OSEr143 seems to be very old, so it could not contain the SMP-related code. Or, probably, your second machine contains cores
with contiguous LAPIC ID's, like I have on my machines, so this error won't happen. Anyway, you can install an OS/4 kernel, and its PSD+ACPI
and we will see if your LAPIC ID's are contiguous. It should write about that in the kernel log.

General Advertisements / Re: OS/2 Discord Channel
« on: December 26, 2019, 12:11:36 am »
Does Discord have an OS/2 client? If not then I see no sense to use it for communication. As well as Telegram, Matrix etc. But Telegram
previously had a web client. Slack had both a web client and gates to irc and jabber. Now these gates are gone away, and web client no
more works with OS/2 browsers. telegram web client too. Though, I'm still using Slack via Bitlbee IRC to different IM networks gateway,
which I run on my Linux VPS server. So, I can connect to it with a simple IRC client from OS/2. Besides Slack, Bitlbee has transports
to ICQ, Jabber, Skype, Telegram, Facebook, VK, Twitter, etc. Of course, voice and video calls are unavailable from IRC, but usually text
is sufficient.

Utilities / Re: Booting OS/2 with an Animation
« on: November 21, 2019, 04:31:52 am »
2Martin: Also, OS/4 supports color logos created as GIF pictures. Not sure if it supports animated GIF's, but everything is possible ;)
Most probably, no, of course ;)

Applications / Re: About bldlevel
« on: September 15, 2019, 04:49:29 am »
2Pete: It seems, bldlevel.rexx requires osFree version of oso001.msg. It is an extension of standard OS/2 version.
You can download oso001.msg, oso001h.msg and oso001.txt (message file source) from

PS: If you don't want to touch your oso001.msg, you can modify the script, so that it will bear messages in its body,
instead of a message file.

Applications / Re: About bldlevel
« on: September 14, 2019, 10:47:43 am »
2Martin: Build Level is just a special string embedded into the binary. Yuri Prokushev researched its format and wrote a doc about it:

Also, he reimplemented bldlevel.exe in REXX:

The buildlevel.exe file you mentioned above is converted from bldlevel.rexx via RexxWrapper (, written by Mark Hessling)

BTW, you can embed the build level string in any file. For example, QSINIT is a plain COM file, but its "os2ldr" contains the build level information. It's just a special string embedded into the binary. AFAIK, in LX/NE executables, it is contained in entry table, if I'm not wrong.

Internet / Re: TOR and FF
« on: August 13, 2019, 02:21:02 am »
2dryeo: I didn't tried HTTP proxies, but socks5 proxies like Tor or SSH tunnel, doesn't work.

You just create a SSH tunnel from local host to

ssh -D 2222 -N

and use it as a Socks 5 proxy on address, and it does not work

The same works fine in FF under Linux. I complained about that to Dmik, he says that it is
some hidden default proxy settings are not correct. But I checked all proxy settings in about:config
and did not found anythin suspicious. All was correct. So, this is indeed Socks proxy settings are broken.
As I said, everything is working fine under Linux. And I don't see why some defaults in OS/2 should be
different than in Linux.

And the same is with SSH and Tor.

Internet / Re: TOR and FF
« on: August 12, 2019, 08:14:36 pm »
2Remy: Unfortunately, it seems, proxies are broken in current versions of FF and SM. So, I won't be able to use Tor or any Http or
Socks proxy in FF / SM. Trying different proxy-related settings didn't help.

Applications / Re: VirtualBox 5.0.51 for OS/2 error
« on: August 01, 2019, 11:35:47 pm »
#define VERR_INVALID_CPU_ID                 (-1018)

This means that invalid CPU ID is specified. On IBM kernels, CPU ID's are calculated as local APIC ID's. Unfortunately, this method
is unreliable, because sometimes CPU core's LAPIC ID's are not from 0..NCpuIDs interval. I think, this is your case. In my case,
my CPU has contiguous interval of LAPIC ID's, so you're unlucky. I'd suspect that you're using IBM's kernel? Unfortunately, IBM
kernels lack a reliable method of getting current CPU ID. OS/4 kernels have a specific KEE API for that. Also, OS/4 kernels have
API for CPU rendez-vous. This is required for enabling hardware virtualization (VT-x/AMD-V). IBM kernels also have no such
feature. So, in your case I'd suggest using OS/4 kernel.

PS: your error occurs when VBox switches to a CPU with a spectfic LAPIC ID. VBox assumes that CPU ID's are consecutive,
but in your case it's not (most probably).

PPS: You also can try pinning VBox executable to a 0 CPU, so it will not hit a hole in CPU ID's range.

Applications / Re: VirtualBox 5.0.51 for OS/2 error
« on: July 31, 2019, 06:54:10 pm »
Did you install VBoxDrv.sys? If yes, is it the latest version?

ZorinOS? What's it? Linux distro? Never used it. Yes, 64-bit guests currently don't work. Yes, some Linux distros
haven't a 32-bit version. The work on hardware virtualization is being worked on, but not yet. It should
work, in theory. At least. I ran Ubuntu 64-bit guest on WinXP 32-bit host, so it can be done on OS/2 too,
but not yet. Why do you need necessarily the exotic ZorinOS, why not e,g, Debian,
which has a 32-bit version?

PS: I have Debian on a bare hardware too (two separate machines and a ssh session from one
to another)

2Doug Bissett: You're right. Another cause of this error could be libcx allocating buffers in high memory. Some OS/2 API's doesn't
like when buffers are in high memory, so they need to use them in lower memory. To use buffers in lower memory, IIRC, LIBCX_HIGHMEM=2
option can be used.

2ivan: No, VboxDrv.sys does not write any messages during the boot. The message you posted can mean different things. It means that it fails
to create the XPCOM object. Something got wrong. The reasons can be different. It usually just fails to load some libraries. Usually, some of the
required libs is missing, or outdated. Or, it fails to loads a library because of a shared memory problems. May also loading libraries to high memory
help. Or, some libraries are missing or outrdated. Are you sure you installed all needed libraries? The download page lists all the dependencies.
You'd better use yum/rpm to install dependencies, because installing everything manually is tricky and not supported.

PS: What VIRTUALADDRESSLIMIT do you have? (it may be too low)

Pages: 1 2 3 [4] 5 6 ... 24