Using BCP to export data from SQL Server 2016 to SQL Server 4.21

Once more again I’m confronted with a situation where I needed a SQL, but I don’t have direct access to the data. The machine I’m able to run some stuff on is not only insanely out of date (yay!) but doesn’t have enough disk space for even something like SQL Server 2000.

Enter SQL 4.21a

I “installed” 4.21a on this 32bit 2003 server in much the same way I transplanted 4.21a onto Windows 10. However I did use the srvany utility to load up the SQL Server service, much like how I used it to run an instance of Qemu in the background elsewhere. Now I have my intermediary SQL Server running like a normal service, and set a password for the SA user.

Now for the fun.

First be sure to set your target database for ‘bulk/load’ and I’d also set it for ‘truncate log on checkpoint’. If you don’t set the bulk/load then you cannot BCP data into the database.

Using the SQL Explorer tool I could view the tables I wanted, and export them as ‘SQL CREATE’ giving me the table layout. I then quickly converted them into something acceptable for 4.21a. Now it’s a matter of establishing a connection to the old server from the new.

First I tested with the ISQL command. I needed to copy the DLLs DBMSSOCN.DLL & NTWDBLIB.DLL into the directory to get the command to fire up. Since my strategy here is to do a BCP dump/BCP load the first thing I need to do is purge the data.

isql  -Usa -PPASSWORD -S192.168.1.42 -Q"delete network.dbo.firewall_mapping"

This of course assumes that the server address is and that in this case I’m deleting the firewall_mapping table from the network database. If you’ve made it this far that means we are 1/4th of the way there!

I found this ‘one trick’ to get the BCP command from the SQL 4.21a tools to connect to the 2016 server and dump the table as a trusted connection. I’m not sure how much longer this kind of thing will work, but I was pretty amazed it did. I didn’t even bother trying to see if the 4.21a BCP tool could read a 2016 BCP dump. Maybe it would if you keep the formatting the same, but I find ‘like to like’ much easier. I renamed the old BCP.EXE to BCP42.EXE so that they won’t collide in any way causing weirdness. At the same time I run them from a directory that is NOT in the system path.

bcp42 "[Network Database].[dbo].[firewall_rules]" out c:\temp\1.csv -t, -r= -P

The notation looks weird, as my source database name has a space in it. This initially caused endless frustration, but it was just a matter of using the fully qualified name, which is in quotes


I set the field delimiter as a comma, and the row terminator as an equal sign. I tried not setting it but I was getting ‘spiraling data’ as it was not picking up the end of row correctly at first.

The first time you run the BCP without a format file it’ll walk you through the specifics of the fields. I just blindly accepted the defaults, and saved the file as firewall_rules.fmt . Now on subsequent runs, I can run the export like this, which uses the saved formatting:

bcp42 "[Network Database].[dbo].[firewall_rules]" out c:\temp\1.csv -t, -r= -P -ffirewall_rules.fmt

Great so if everything is going well, we have no exported our data! Now the next step is to import the data into our old server. Since we have that format file, this “should” go pretty smoothly. Notice the server is an IP address which implicitly has it connecting by TCP sockets, not named pipes. As such there is no implicit ‘Trusted connection” as there was when talking to the local 2016 server.

bcp42 network.dbo.firewall_rules in c:\temp\1.csv -Usa -PPASSWORD -S192.168.1.42 -t, -r= -ffirewall_rules.fmt

Naturally change PASSWORD to whatever password you have for the SA user.

Starting copy...
1000 rows sent to SQL Server.

1892 rows copied.
Network packet size (bytes): 512
Clock Time (ms.): total =   2216 Avg =      1 (853.79 rows per sec.)

And there you have it, all being well you’ll see the program update every 1,000 rows as it inserts data.

Originally I wanted to use the data transformation wizard thing (whatever they renamed DTS to) however the ODBC is limited to the newer .NET 4 stuff, which won’t use the old SQL Server 6.5 ODBC drivers. I really didn’t think the SQL Server 4.21a BCP command to work on a modern server against a new(ish) version of SQL Server, but it did!

I guess you could neaten it up with a command file to drop tables/re-create if you wanted, or at the least delete data/checkpoint and set the load options, dump/load data, and then turn off the load state for the database. I’m not doing reports or anything fancy, just visualizing data as they say.

Although things like ODBC have drifted out, it’s still kind of interesting that ancient BCP can still communicate over named pipes as an implicit trust.


(This is a guest post by Antoni Sawicki aka Tenox)

As I previously mentioned I’m slowly but surely uploading all my stuff to

You may want to know that I just finished uploading apps for Windows NT RISC… Alpha AXP, MIPS and PowerPC. Happy downloading.

Early MSDN CD’s on

I ran across this, and thought it was cool. These CD’s are getting harder and harder to find, and unless you want the old physical disks, getting ISO images is, of course the next best thing.

Granted these disks replaced the much older Microsoft Programmer’s Library. The new CD’s use a Windows based search & interface program removing the clunky old MS-DOS program that made it feel like trying to view the world through a straw. (Although the up side of the MS-DOS version is that you could easily dump the video RAM and save the contents to plain text).

And in this brave new post Windows 3.0 centric world of Microsoft just about everything regarding OS/2 was dumped, and the seeding of Win32 via Windows NT had started.

Naturally after winning this war, Microsoft withdrew many low end products and just couldn’t compete with the tidalwave that was GNU/Linux.

At any rate for the curious kids down the road that want to see what all the fuss was with Win16, and how Windows 3.0 had changed the landscape removing the force of IBM it’s worth a look.

Installing Windows 10 pro onto an older Mac Pro

So I have this 2006 Mac Pro 1,1 that I’ve had laying around and I wanted to put my old Nvidia 1030 into it, along with Windows 10 for a newer (stronger?) home machine.

So I burnt the downloadable ISO from Microsoft onto a DVD, tried to boot it up and got this:

Ignore the microwave…

I got stuck at this “Select CD-ROM Boot Type : ” prompt, which you can’t type anything into. Apparently it’s a common and known issue with 64bit boot code, as the older Intel Apple Mac’s are of course 32bit only. So there is a fix, you have to use something called “oscdimg” to rebuild the ISO with a 32bit friendly loader.

So first I just used 7zip to extract the downloaded ISO, and then create the new 32bit ISO with the following:

oscdimg -n -m -bc:\temp\Windows10\boot\ c:\temp\Windows10 c:\temp\LegacyBootWindows10dvd.iso

And yes, it boots!


Now what about those special ‘bootcamp’ drivers? Obviously under Windows 10 there is no nice way to get the ones you want. And what about which model/download package to get?

Then I found Brigadier!

It’ll probe your hardware type and grab the drivers for you!

So I picked up this Nokia 1020, and decided to downgrade it to Windows 8.0

When I’d bought this 1020 in question, I picked it up from an AT&T customer with the intention of basically using it as a camera.  Well it was kind of okay but it had been upgraded to Windows 10, which technically is unsupported on the 1020 hardware.  And no wonder as the 1020 specific camera applications aren’t available, nor is performance all that great on Windows 10.  I’m guessing there is a laundry list of reasons of why Windows 10 was not available for the 1020.

So I saw this website, that not only has an incredible amount of firmware saved, but it sure makes the process of loading different firmware pretty simple.

  • Install “Windows Device Recovery Tool”
  • Download the firmware form this site only the ffu.
  • And now open CMD as administrator
  • Type “cd C:\Program Files (x86)\Microsoft Care Suite\Windows Device Recovery Tool”(32 bit pc) “cd C:\Program Files\Microsoft Care Suite\Windows Device Recovery Tool”
  • Connect now your phone to your pc
  • Type in cmd “thor2 -mode uefiflash -ffufile “C:\the location of the ffu\file name.ffu” -do_full_nvi_update -do_factory_reset”
  • Wait now until your phone gives a green screen
  • Type now in cmd “thor2 -mode rnd -bootnormalmode”

Really doesn’t that sound not that difficult at all?!

Except it wasn’t.

Looking at the incredibly extensive list, I see that the AT&T 1020 phone in yellow colour is the RM877.  So I download the appropriate ffu image and run:

thor2 -mode uefiflash -ffufile RM877_3051.50009.1451.1_RETAIL_nam_att_205_01_473928_prd_signed.ffu -do_full_nvi_update -do_factory_reset

A bunch of verbose output happens and then:

Start sending payload data V2Sbl in async mode...
Percents: 0
Device responded a message that has invalid size!
Expected: 16, Received: 0
Exception during programming: 262150
Safe write descriptor index reached: false
Payload data transfer speed (0.22 MB/s) Elapsed time 40.83 sec
Payload data size 9.000123 MB
[IN] programSecureFfuFile. Closing RM877_3051.50009.1451.1_RETAIL_nam_att_205_01_473928_prd_signed.ffu
programming operation failed!
Operation took about 1 minute, 6 seconds. Average transfer speed was 0.11 MB/s.

Unknown error code.

THOR2 exited with error code 262150 (0x40006)

Unknown error code!  What the hell does that mean?

Glancing over at the phone I’ll show you what it means.

Nokia Red screen, waiting to be programmed

So the phone was stuck at the red screen waiting to be programmed, but obviously that failed.  Re-running the programmer just gave me this:

Detecting UEFI responder
Lumia UEFI Application did not respond to version info query
Device is not in Lumia UEFI mode
Reading device mode failed
Failed to detect UEFI responder. 84017


THOR2 exited with error code 84102 (0x14886)

So I thought I’d hold down the volume down button & power and do a reset.  Except now my device was nearly bricked.  The display wouldn’t turn on at all, and there was no haptic feedback on powering up and or down.

HOWEVER plugging in the USB would make the connection noise, and how me a “QHSUSB_DLOAD” device.  So I could try to talk to it in emergency mode.

Apparently on the same download page, there is a hex file and a mbn file.  Downloading those and running:

thor2 -mode emergency -hexfile FAST8960_EOS_NAM.hex -mbnfile RM877_msimage_v1.0.mbn -orig_gpt

And this looked like it was going to work… but then as always it failed.

Received valid response to OPEN_MULTI_REQ
Programming image R
Image opened successfully for reading
SAFE hex file was used and unallowed memory address was being written.
Reset the device and use the correct HEX file.
Emergency messaging closed successfully
Operation took about 11.00 seconds.

Unknown error code.

THOR2 exited with error code 85034 (0x14C2A)

Use the correct HEX file? But it was the one that I downloaded from the right page!  Great now my phone is dead.  Unplugging it in and out however gives me a little hope as it’s still showing up as a QHSUB_DLOAD device.

After a lot of searching I see Joemar Serrato went down the same road, and they did something weird where they split out the mbnfile from the ffu file themselves.  So I ran the following command to split out the mbn files:

thor2 -mode ffureader -ffufile RM877_3045.0000.1325.0001_RETAIL_nam_att_205_01_223461_prd_signed.ffu -dump_gpt -filedir gpt

And now I had 2 files.

10/04/2018  06:53 PM            17,408 GPT0.bin
10/04/2018  06:53 PM            17,408 GPT1.bin

Now with all that drama I could re-run the emergency mode flash

thor2 -mode emergency -hexfile FAST8960_EOS_NAM.hex -mbnfile gpt\GPT0.bin -orig_gpt

Received valid response to OPEN_MULTI_REQ
Programming image g
Image opened successfully for reading
Uploading MBN image 70
SAFE hex file was used and unallowed memory address was being written.
Reset the device and use the correct HEX file.
Emergency messaging closed successfully
Operation took about 11.00 seconds.

Unknown error code.

THOR2 exited with error code 85034 (0x14C2A)

And it appeared nothing changed at all. Great.  In disgust I unplug the phone and hold down the volume down & power button for 15 seconds, but this time it vibrated.  I couldn’t believe it!  Within a few seconds it powered back up to the red screen, ready to load new firmware!

thor2 -mode uefiflash -ffufile RM877_3045.0000.1325.0001_RETAIL_nam_att_205_01_223461_prd_signed.ffu -do_full_nvi_update -do_factory_reset

Size of system mem: 2097152 KB
Send backup to RAM req...
Clearing the backup GPT...SKIPPED!
Successfully parsed FFU file. Header size: 0x000c0000, Payload size: 0x00000000572e0000, Chunk size: 0x00020000, Header offset: 0x00000000, Payload offset: 0x00000000000c0000
RKH match between device and FFU file!
Option: Skip CRC32 check in use
Start sending header data…
Start sending payload data V2Sbl in async mode...
Percents: 0
Percents: 1
Percents: 2
Percents: 3
Percents: 4
Percents: 5

I was astonished!  The progress bar was slowly moving, and the phone was doing what it should have been doing about 2 hours ago.  I dared not move, so unfortunately I don’t have any pictures.

Payload data transfer speed (9.79 MB/s) Elapsed time 142.41 sec
Payload data size 1394.893921 MB
Read flashing status..
[IN] programSecureFfuFile. Closing RM877_3045.0000.1325.0001_RETAIL_nam_att_205_01_223461_prd_signed.ffu
Get EMMC write speed...
Get EMMC write speed, SKIPPED!
Get data verify speed...
Get data verify speed, SKIPPED!
Send restore backup from RAM req...
programming operation completed!
[THOR2_flash_state] Post programming operations
[THOR2_flash_state] Executing factory reset
[Factory reset result] 0
Factory reset done
[THOR2_flash_state] Executing Full NVI Update
Write flash options ( WriteNvi: Full )

Write parameter to device via UEFI Flash App
Write parameter Ok
Flash options set successfully.

[Full NVI update result] 0
Operation took about 2 minutes, 24 seconds. Average transfer speed was 10.16 MB/s.

Exited with success

And then just like that, the phone rebooted itself, and then brought itself up to the ‘green screen’

Nokia green screen (firmware loaded successfully)

All that remained to do was to tell the phone to boot up normally

thor2 -mode rnd -bootnormalmode

Version 8.0.10327.77

And that was it!

My 1020 is now running 8.0, which includes not only the special Nokia phone apps, but all the AT&T bloatware of 2012.  I can’t believe there was a yellow pages app.  How are they still a thing?

Wileyfox Windows phone back in production!

And people think Amiga & OS/2 fans are a little crazy!

Wileyfox was the last vendor selling Windows phones and it seems that they there was so much demand that the phone is back into production.

The WileyFox Windows phone

Specs include:

• 5″ HD Curved Screen, IPS Technology, Gorilla Glass 3
• 4G LTE
• Windows 10 Operating System
• Fingerprint Reader
• 8 Megapixel Rear Camera + Dual LED Flash + Auto Focus
• 2 Megapixel Rear Camera
• Qualcomm Snapdragon Quad Core Processor
• 16GB ROM (Storage)
• 32GB Expandable Memory Card Slot
• 3D G Sensor
• Proximity Sensor
• Ambient Light Sensor

Also the price has dropped from £ 199 to £ 79!

Currently it seems that orders are only for the United Kingdom.

Loading the MS SQL 6.5 drivers on Windows 10

AKA accessing SQL Server 4.21 via Access 2016

It actually works!

I’ve been pretty dismayed for the longest while that newer versions of Windows bundle newer versions of ODBC that refuse to talk to any version of SQL Server prior to 2000 (at the moment).  Of course if I were ‘professional‘ Id be upgraded to the latest version, maybe even running on Linux.  But I’m not, and you didn’t come here for how to upgrade/update but rather how can you use tools from the 90’s in the 10’s.

If you try to use the current SQL Server driver, you’ll get this message.  Don’t be fooled, SQL Server 7.0 isn’t supported either (probably because of the SUN vs Microsoft Java debacle), it will only work with SQL Server 2000 or later versions.

SQL Server version 6.5 and all previous versions are no longer supported

It’s been this way since Windows 7, and I’ve just given up and gotten used to having to have a VM to access older databases.  That is, until today.

While I was dumping data and moving stuff for my blog (something I need to write about another BCP adventure), I wanted to do a simple Access database to make sure it’s looking sane.  And I figured I’d jump down to Windows 95, and load up Office 95.  Well sure enough there was no proper SQL Server ODBC driver.  Popping in the SQL Server 6.5 CD, naturally there is no ODBC drivers for Windows 95, but rather a 16bit driver for Windows 3.1, and 32bit drivers for Windows NT.  I installed the NT version, and was still unable to connect until I installed the SQL client which then let me run WINDBVER.EXE so I could configure the appropriate transport DLL (TCP/IP) and then I could connect.

Access 95 using the SQL 6.5 NT ODBC Driver + Client

And this got me thinking, is it possible to just take those DLL’s and move them onto Windows 10?  Well naturally that won’t work as the driver sqlsrv32.dll is a system protected file, and you can’t overwrite it without a lot of pain.  However this got me to thinking that it should be possible to just tell the system it’s a new driver with a different name.

Looking through the registry I notice that HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\ODBC is where the 32bit ODBC stuff is stored, so this is where I’ll make my ‘OLD SQL’ driver.  I simply used the existing one and renamed the driver/setup DLL’s to sql65-sqlsrv32.dll so that way the 2 DLL’s can co-exist without freaking out.

OLD SQL Driver

With that done, the key Computer\HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\ODBC\ODBCINST.INI\ODBC Drivers will need a new SZ string saying the ‘OLD SQL’ driver is installed.

Add the new driver to the available list.

I also copied the following files from the SQL 6.5 client into a subdirectory of SysWOW64 so I could configure the transport:


Now if you are going to use named pipes it’ll just work out of the box.  Running WINDBVER you’ll see that it’s set that way by default.

Named pipes, and notice it’s version from 2016!

to change it to TCP/IP I coped the files DBMSSOC3.DLL & DBMSSOCN.DLL into the SysWOW64 directory and re-ran WINDBVER, and then selected TCP/IP.


Now we can go ahead and add a SQL DSN. Remember to use the 32-bit ODBC Data Source tool

Note this is the 32-bit version!

 If everything is setup correctly you should see the ‘OLD SQL’ Database driver.  Scrolling to the right you’ll see that it’s version dated 7/7/1997

Use the ‘OLD SQL’ driver

Now we just need to configured the driver.  Be sure to hit the Options button and type in the database name, otherwise it’ll just go to the default database.  Since I’m just using the ‘sa’ user that would be the master database, but I want Access to instead use the ‘pubs’ database.

Configure the old driver

Sadly these old drivers don’t have any test functionality.  So there is no real way to know if it is working at this point, however I would imagine it should as we have added a regsitery key for the driver, added it to the available driver list, selected (if needed) a transport DLL, and populated the needed fields.  The next thing to do is to try to use it.

The ODBC option is in the ‘Other’ sources now.

Start Access, and select a blank database.  Go to the External Data tab, and then choose the New Data Source, and the From Other Sources option.  This will bring up the ODBC database wizard.

Link to the data source

I’m going with linked, as I want the data to be used from the SQL Server all of the time.  You will be then prompted to choose the Data Source

The dbtest ODBC connection should be under the Machine listings

You can select the DSN we setup earlier, and then it’ll prompt you to login.

Login to the SQL Server

Again I’m using SA because… why not.  And now for the moment of truth, if everything is correct you’ll be presented with the list of tables!  This means that it’s working!

The pubs database from SQL 4.21

You can choose to save the password, and if needed select unique fields, or just leave it as it is.  Now you should be able to access your SQL 4.21 database!

The Authors table from the pubs database

And there we have it.

If anyone is brave/crazy enough I extracted the driver here: 
sql65odbc-files.7z  SQL Server 6.5 should not only be able to access 4.21, and 6.0 but I think it ought to be able to access SQL Server 7.0, although I haven’t tested it out yet.

OpenNT 4.5 revisited

OpenNT 4.5 installed

It’s been a few years since Tenox had mentioned OpenNT 4.5, and in that time the project pages, repositories and well just about everything has vanished.  It seems that the hardest thing to do with OpenNT had become finding it.

Then I found this over on vetusware, and with my curiosity piqued, I thought I’d check it out.

As mentioned the first thing to do is combine the parts, and create the single 7zip file, and then extract that.

copy /B "OpenNT45(Source).z00" + "OpenNT45(Source).z01" + "OpenNT45(Source).z02" "OpenNT45(Source).7z"

Extracting that will give a simple ISO file, weighing in at 1.7GB!

While not obvious at first, there is a readme in the ISO that provides instructions on how to compile it.  Basically it boils down to a few main points:

  • xcopy the CD onto a ‘W’ drive, or subst any other drive to ‘W’ as apparently the build process requires it to be on W.  Did I mention that the CD needs to be copied onto the W drive?
  • Run the ‘setup.cmd’ file to configure the environment and get the build process ready
  • run zTESTBUILD and do a clean build.  It will run and eventually fail.
  • run zTESTBUILD again, but do not do the clean build, and it should finish
  • run \cdimg\genall to create the ISO image

So with those points basically figured out after the fact, let’s go!

The first thing to do is either create a VM to compile this in, or just xcopy and go.  The big requirement though is that it must be a 32bit version of Windows, as part of the build process requires the ability to use NTVDM.  For simplicity sake, I chose Windows 2000 server, so I could allocate 2GB of RAM, and 4 CPU cores.  During the build it doesn’t use that much memory, cores are more so important during various phases of the build that can seemingly use any and all cores, while various other parts only run on a single core.

OpenNT disk space after a build.

I chose to use a separate ‘C’ drive, and ‘W’ drive for the 2000 VM.  With no idea how much space to give it, I setup a 32GB W drive, which after the build takes up just under 4GB of space.  

With the VM installed, and the W drive formatted, and the contents of the CD copied over, it’s time to start the build.

Phase 1

So basically you just answer ‘Y’ to zTESTBUILD.cmd and it’ll do it’s thing.  For me this took about 42 minutes for this to run until it failed, as expected.

Failure building MVDM

Looking at the \binaries\nt directory there was now 1,274 files currently built.  Naturally with the failure this is not a complete build.

Starting phase 2

After this failure you then re-run zTESTBUILD.cmd but this time answer ‘No’ to the clean build.

This step took about 15 minutes to complete.

The \binaries\nt directory now has 1293 files.

Checking the binaries\nt directory there is now 1,293 files and looking at the entire directory there is 2,624 files taking up about 120MB of space.

With the OS compiled, all that remains is to create an install CD and boot it up.  running \cdimg\genall.cmd will create the ISO image.

Creating the ISO with CDIMAGE 2.47

This will compress almost all the files, and took another 15 minutes to create the CD.  After this is all done it’s just a matter of setting up a VM to run the NT45Wks.iso file.

Bluescreen bootup of OpenNT 4.5

The first thing you notice is the extra banners on OpenNT 4.5, when compared to a retail copy of NT 4.0.

OpenNT Project

And of course the different branding during setup.  One of the nice things about OpenNT is that it can format filesystems directly as NTFS, instead of the old way of first creating a FAT partition, and converting it to NTFS.  This ought to bypass all the limitations of disk/partition sizes for the older NT.

WOW Crash

Running OpenNT 4.5 on VMWare seemed to run the Win32 stuff okay, although Win16/WOW stuff immediately crashed, and MS-DOS was incredibly slow with screen redraw issues. I know that NT 4.0 builds prior to SP 6 have issues with many newer emulation/hypervisors even when CPU levels are set to regular Pentium.

MS-DOS DPMI stuff like DooM are incredibly slow, and seemingly lock up when launching.

Oddly enough the OS/2 subsystem works just fine.

So there it is, the nearly vanished ‘OpenNT 4.5’.

VMware Player and Device/Credential Guard are not compatible!

What is this?!

Well it turns out that by turning on all the SDK stuff for Windows 10, including the mobile dev, which includes the Windows phone emulator it naturally uses Hyper-V.

Hyper-V Hypervisor enabled

And obviously the two hypervisor’s wont play nice with each-other.  You could just disable it, and go back and forth re-enabling it when needed, or make a new boot selection without it!

I found this post here: Switch Easily Between VirtualBox And Hyper-V WithA BCD Edit Boot Entry In Windows 8.1

C:\> bcdedit /copy {current} /d "No Hyper-V" 
The entry was successfully copied to {ff-23-113-824e-5c5144ea}. 

C:\> bcdedit /set {ff-23-113-824e-5c5144ea} hypervisorlaunchtype off 
The operation completed successfully.

note: The ID generated from the first command is what you use in the second one. Don’t just run it verbatim.

When you restart, you’ll then just see a menu with two options…

  • Windows 10
  • No Hyper-V

Sounds easy enough, doesn’t it?

BCD Boot menu

And just like that on power up, I can switch between Hyper-v and no Hyper-V.

Oh yeah with the latest version of Windows 10 (October 2018) I had to list the BCD table with:

bcdedit /enum ACTIVE /v

As after the upgrade it had tagged both of my boot selections to enable Hyper-V everywhere. I had to delete the #2 entry and re-create the no hyperv dance. {current} doesn’t work anymore.