Hi Erik,
thanks for your ideas - based on that I made some more tests with my 'Warp5 installation' on HPFS partitions with amaizing results:
1. the COMP-error does not directly depend on file size. Checked with large (10MB) and small (50kB) files:
The chance to get that error is about 50% independent of file size.
2. Verifying the original and copied file by md5sum tells: both files are identical! (though COMP reports diffs).
3. The executable 'COMP.COM' in my Warp5 is the very same as in eCSrc7: 35,881 Bytes, md5sum ......a5f1
4. I payed with the CACHE of HPFS - even with no cache (/CACHE:0) the errors persist.
5. I waried RAM for my Warp5 from 48MB - 1GB: no difference.
6. I did copy the file, verify it by md5sum and shut down. After reboot COMP tells they are differen???
7. I downgraded HPFS.IFS from BldLevel 14.093 to 14.056 (as in MCP): the same.
8. I tested all with a fixed size *.vdi - so disk does not matter
9. I tested with a fixed disk in VMDK format, no snapshots - the same
10. I tested different filnames for source and copy - no difference
11. I did compare by SHA-1 hash (instead of md5sum) the same = identical
What I do observe is: the copying the file happens immediately, while a COMP takes 10 sec. or even more?
(md5sum takes a second, not more)
I verified my whole installation:
I have tar-archives of each partition on a DVD-ROM for quick and easy restore (in case of problems).
The DVD can boot, contains some Tools like FDISK, FORMAT, TAR.
Of my whole virgin installation from that DVD (extracted tar-archives) I immediately made a snapshot. After reverting to that snapshot and booting from DVD I made a tar-compare (tar -dppP -f <archive>) of all partitions - with absolutely ZERO diffs!!!
So the whole story is really obscure:
copying several GB of data and verifying with tar is fine, copying single files within OS/2 is fine, verify by md5sum is fine - just COMP fails (the very same as you have)
This COMP-bug even survives shutdown of the VM, powering it off, closing the VBox application, and even reboot of the host (Ubuntu-Hardy-amd64).
After reboot it is the same story.
So the only things remaining are
HPFS itself and my partition scheme
---------------------------------------------------
Finally I just found a workaround:
1. copy the file (say uniaud32.sys) to another place
2. just calculate md5sum of the
source-file (not the copy)
3. now COMP.COM reports correctly identical files - even after subsequent copies or whatever!
If you have rebooted the VM (or shutdown+power off inbetween), it is necessary to calculate md5sum of both files, source and target to make COMP happy. Doing that with only one of them is not sufficient.
Any explanations ??? To me it is
mysterious
Ingo