issinoho wrote: ↑Sat Jan 25, 2025 5:26 am
So given all the above, I'm not inclined to try to change anything here. There doesn't appear to be an easy way to break the alias (am I wrong?) leaving the original files intact, so we are where we are and I'm just going to leave the SYSLOST files well alone.
Leaving it like it is may be ok for the short term, but it isn't good to leave it like it is for the long term.
This is especially true if you use purge, because OpenVMS (even 8.3 and 8.4) don't handle this in the way I would consider correct. At least it is very confusing. And a test i just did on a small LD device with ODS-5 and hard links enabled showed the same results as when using an ODS-2. John Briggs showed an example of this in this comp.os.vms thread Deleting alias files (blocks deleted)
https://groups.google.com/g/comp.os.vms/c/gLqmWwoXqZI
The most interesting part of the thread starts with Hein's post from Dec 20, 2005, 11:05:50 PM here:
https://groups.google.com/g/comp.os.vms ... vTAtpRTrcJ
John Brigg's two examples start here: (this and the following post)
https://groups.google.com/g/comp.os.vms ... v6niXQRR0J
My recommendation would be to fix, but in your case manual intervention will probably be needed.
I would also recommend creating a couple of small 1000 block LD disk ($ LD help setup), initialize one with /struc=2 and the other with /structure=5/volume=hard, and then play with set file /enter and set file /remove (and delete and purge)
Here's a thread on how a common case of incorrect backlinks (vms$common.dir vs syscommon.dir) can be fixed.
https://forum.vmssoftware.com/viewtopic ... 1084#p2659
Oh, and I found this on bitsavers
https://bitsavers.org/pdf/dec/vax/vms/t ... s_1990.pdf not light reading, but much easier to read than the Runoff for hardcopy printers version of the ODS-2 spec from 1985, which formed the basis of chapter two of EY-F575E-DP.
Here's a comand file I used to reproduce John Brigg's results, TEST_ALIAS.COM
Code: Select all
$ ver = 'f$verify(1)'
$ set noon
$! TEST_ALIAS.COM
$! Jon Pinkley - 2025-01-30 response to Understanding SYSLOST
$! see https://forum.vmssoftware.com/viewtopic.php?f=31&t=9364
$! demonstrate inconsistent behavior with alias entries, delete vs purge (based on John Brigg's posts comp.os.vms)
$! for more info, see these old comp.os.vms threads
$! Deleting alias files (blocks deleted) https://groups.google.com/g/comp.os.vms/c/gLqmWwoXqZI
$! ODS5 and hardlinks https://groups.google.com/g/comp.os.vms/c/TQhvQgqH5hs
$ this_proc=f$environment("procedure")
$ proc_dir = f$parse(this_proc,,,"device") + f$parse(this_proc,,,"directory")
$ orig_default = f$environment("default")
$ show symbol orig_default
$ show default
$ show symbol this_proc
$ show symbol proc_dir
$ set default 'proc_dir'
$ show default
$
$ this_pid = f$getjpi("","PID")
$ cnt = 0
$
$_credir_top:
$ dir_str = "TEST_''this_pid'_''cnt'"
$ if f$search("''dir_str'.DIR") .nes. ""
$ then
$ write sys$output "''dir_str'.DIR already exists, skipping"
$ cnt = cnt + 1
$ goto _credir_top
$ else
$ write sys$output "Creating [.''dir_str']"
$ create/dir [.'dir_str']/version_limit=2/log ! create a new directory, and set default version limit to 2 so new files created wi
ll have version limit in directory
$ endif
$
$ set default [.'dir_str'] ! make new subdirectoy the current work directory (default directory)
$ pipe dir/ful [-]'dir_str'.dir | search* sys$pipe " version"
$ dump/dir [-]'dir_str'.dir ! dump empty new directory
$ create a.a;1
$ DECK
a.a;1
$ EOD
$ set file/ent=b.b;1 a.a;1 ! create an alias for a.a;1 called b.b;1
$ dir/file
$ type a.a;1
$ type b.b;1
$ dump/dir [-]'dir_str'.dir
$! try to delete the alias
$ delete b.b;1/log
$ dir/file/date=(cre,mod)
$ dfu dir/dump [-]'dir_str'.dir
$ type a.a;1
$ type b.b;1
$! in this case it just removed the entry from the directory.
$ set file/ent=b.b a.a ! create an alias for a.a;1 called b.b;1
$ dir/file/date=(cre,mod)
$ create b.b;2 ! b.b;2
$ DECK
b.b;2
$ EOD
$ dump/dir [-]'dir_str'.dir
$ type a.a;1
$ type b.b;1
$ type b.b;2
$ pipe dump/head/bl=c:0 a.a;1 | search* sys$pipe "file ident","file name:","dump of"
$ pipe dump/head/bl=c:0 b.b;1 | search* sys$pipe "file ident","file name:","dump of"
$ pipe dump/head/bl=c:0 b.b;2 | search* sys$pipe "file ident","file name:","dump of"
$! try to delete the alias to see if having another version of the file makes a difference
$ delete b.b;1/log
$ dir/file/date=(cre,mod)
$ if f$search("a.a;1") .nes. ""
$ then
$ ! still did not delete the primary (a.a;1), just removed alias directory entry.
$ ! Add alials back again and see if purge does same thing; it is using the same image (DELETE).
$ set file/rem b.b;1 ! should already be gone
$ set file/ent=b.b;1 a.a;1 ! create an alias for a.a;1 called b.b;1
$ endif
$ dir/file/date=(cre,mod)
$ purge b.b/log
$ dir/file/date=(cre,mod)
$! but it does not, it deletes the primary entry and the file. I consider this a bug, but it won't be fixed.
$! Aliases should be used only with great caution.
$ if f$search("a.a;1") .eqs. ""
$ then
$ ! recreate a.a;1 and alias b.b;1
$ create a.a;1
$ DECK
a.a;1
$ EOD
$ set file/rem b.b;1
$ set file/ent=b.b;1 a.a;1 ! create an alias for a.a;1 called b.b;1
$ endif
$ dir/file
$ type a.a;1
$ type b.b;1
$ type b.b;2
$ dump/dir [-]'dir_str'.dir
$!now create a third version of b.b, which will invoke auto purge to 2 versions of b.b in the test.dir, but this is an "alias" and n
ot primary.
$ create b.b; ! b.b;3
$ DECK
b.b;3
$ EOD
$ dump/dir [-]'dir_str'.dir
$ type a.a;1
$ type b.b;1
$ type b.b;2
$ type b.b;3
$_cleanup:
$ set default 'orig_default'
$ show default
$ exit 1+0*f$verify(ver)
Here's another thread on the old HP ITRC forum concerning what was apparently a system disk corruption due to a partitioned cluster. Hopefully you did not have any "%ANALDISK-W-MULTALLOC, file (file-id) filename.type;ver
multiply allocated blocks" errors, as those are a sign of serious corruption.
https://community.hpe.com/t5/operating- ... -p/4265788
