F$DEVICE lexical

[Art Wiens]

If one reads:

Lexicals
F$DEVICE
Arguments
search_devnam

Specifies the name of the device for which F$DEVICE is to search. The asterisk (*) and the percent sign (%) wildcard characters are allowed in the search_devnam argument.

Specify the search_devnam argument as a character string expression.

Why would you not be able to use:

$ disk = f$getdvi("$1$DK*","DISK")

It works if you specify f$getdvi("*","DISK"), but I only want info on the scsi disks with allocation class $1$.

* isn't _really_ a wildcard?

Cheers,
Art

[Art Wiens]

I'm having a bad day here...one more time.

The f$device lexical help reads that you can use an * wildcard...I made a slip in my post (f$getdvi), but in my command procedure I _am_ using f$device and it doesn't work.

Maybe one too many "Cheers" ;-)
Art

[Lawrence Czlapinski · Accepted solution]

Art,
1. You need to specify device type which is usually "GENERIC_DK".
2. Unfortunately the "*" has to be a prefix.
CNSRV1» device_name = f$device("*0:","DISK","GENERIC_DK")
CNSRV1» SH SYM DEVICE_NAME
DEVICE_NAME = "_$1$DKA400:"
CNSRV1» device_name = f$device
("*0:","DISK","GENERIC_DK")
CNSRV1» SH SYM DEVICE_NAME
DEVICE_NAME = "_$1$DKB100:"
CNSRV1» device_name = f$device("*00:","DISK","GENERIC_DK")
CNSRV1» SH SYM DEVICE_NAME
DEVICE_NAME = "_$1$DKA400:"
CNSRV1» device_name = f$device("*:","DISK","GENERIC_DK")
CNSRV1» SH SYM DEVICE_NAME
DEVICE_NAME = "_$1$DKA400:"
CNSRV1» device_name = f$device("*100:","DISK","GENERIC_DK")
CNSRV1» SH SYM DEVICE_NAME
DEVICE_NAME = "_

[David B Sneddon]

Art,

The following works for me...

$ say f$device("$255$DK*:","DISK")
! nothing found on this one

$ say f$device("*DK*:","DISK")
_$255$DKB0:

$ say f$device("_$255$DK*:","DISK")
_$255$DKB0:

Not quite sure how the matching is done, I will
have a look at the source listings.

Regards
Dave

[Wim Van den Wyngaert]

I also found that DISK is not a correct device class (http://www.sysworks.com.au/disk$vaxdocmar963/decw$book/dy4yaaa6.p130.decw$book#2390).
It should be DC$_DISK.

And the device type should be one of the list http://www.sysworks.com.au/disk$vaxdocmar963/decw$book/dy4yaaa6.p131.decw$book#2392

Confusing.

Wim

[Karl Rohwedder]

F$DEVICE("*$1$DK*","DISK") works perfectly on our system.

[Wim Van den Wyngaert]

Karl,

I agree because I use disk too. But it's not the option that is documented.

Wim

[David B Sneddon]

From Karl's example and mine above it would
appear to have something to do with the leading
underscore character. "_$1$DK*" works and
"*$1$DK*" works but "$1$DK*" doesn't.
Haven't had a chance to check the source listings
yet.

Dave

[Willem Grooters]

Working on a not-clustered VMS7.3-1 system, I found that this works:

f=f$device("*$DK*", "DISK")

Each call will yield the next device (in arbitrary order as the HELP tells me).

So if you just want the $1$-devices, put a * in front and after:

f=f$device("*$1$DK*", "DISK")

may give the required result.

But it isn't very logically you seem to need the both....

HTH

Willem

[Dale A. Marcy]

It appears that it is doing the comparison to the physical device name which always contains a leading underscore. If you include the underscore or wildcard it, it will match. If you notice all of the values returned, if a match is found, contain the leading the underscore, so I think this is what it is matching the wildcarded value against.

[Lawrence Czlapinski]

Wim, DC$_ is the prefix for device class. The device class names are the part after the DC$.
Art,
ASDEV1» device_name=f$device("*$4$*:","DISK")
ASDEV1» sh sym device_name

In a DCL procedure you can do something like:
$ class = 1
$ START:
$ DEVICE_NAME = -
F$DEVICE("*$''class'$:","DISK","RA60")
$ IF DEVICE_NAME .EQS. "" THEN EXIT
$ SHOW SYMBOL DEVICE_NAME
$ GOTO START

Lawr

[Wim Van den Wyngaert]

I would say the doc need a review.

In 7.3 : f$dev points to doc of f$getdvi. There it points to examples that don't exist. So no list of allowed values.

If DC$ is a prefix, it should be documented.
If _ is required, idem.

Since the beginning (87) I have the impression that each VMS OS programmer was free to implement devices his own way (: or not, _, prefixes etc).

Wim

[John Yu_1]

Seems to me that
f$device("*$1$dk*:","DISK")
should work fine.

[Hein van den Heuvel]

The search name is used with STR$MATCH_WILD.

If any help needs to be improved it is HELP RTL STR$ STR$MATCH because it does not even mention the % and * characters.

Device names do contain a leading underscore which stops logical name translation. That's just how it is. Documented and readily visible with the "*" example

It would seem clear to me that "_xxx" does not match "xxx" but "_xxx", "*xxx" and "%xxx" do match that.

fwiw,
Hein.

[Art Wiens]

Ten points for everyone because it's Christmas! ;-).

Thanks all, I understand now.

Sorry for the delay on points assignment, it's been a "scary" time...a remote VAX (production critical!) working away for the past 10 years had a system disk failure. No problem ... we'll just get another disk and restore the most recent backup tape. What's that, you stopped doing backups "a long time ago"?!?!?!?!? :-O

I sent them two disks with v6.2 on them, they mixed up which were mine, which was their old disk! :-O

They've managed to find some 8mm's from "a long time ago". I'm giving them DCL backup commands over the phone, they're making typo's and typing in things I didn't tell them to!

First tape gave parity errors right off the bat. Second one's running right now. Hard to say how it will end...

No cheers right now, but thanks,
Art

[Jan van den Ende]

Art,

let me wish you strength!

And then, if you have wrestled though this, then rhere will be some teaching, I hope?
And some VERY plain, VERY stupid, VERY short, but VERY consise instructions!

and remember: "may you live in interesting times" is more a curse than a good wish.


Cheers.

Have one on me (and if you get things going again, that IS a deserved one)

Jan

[Uwe Zessin]

Prefixing a device name with '_' to stop any further translation is very old and even predates the TERMINAL logical name attribute, which, if I recall correctly, was introduced with VAX/VMS V4.0 when the logical name system was overhauled.

[Jan van den Ende]

Uwe,

I don't remember the introduction of /TERMINAL, but I _DO_ know it became much less relevant, to the extend that I consider it obsolete, at V5.1 (maybe 5.0, never had that on my hands, nor 6.0 nor 7.0, for that matter)
Under V4.x (and I -think- V3, but unsure)
Concealed Logical Names had to be /TERMINAL, and so they had to reference the physical device directly.
But since V5, if you simply omit /terminal, everything works fine, and you can point a concealed device to a directory structure on the concealed logical name for a disk, that does not need to be terminal, and may even translate to another conc. lnm for the disk.

Makes things more simple, which (to me) is always welcome.

hth

Cheers.

Have one on me.

Jan

[Uwe Zessin]

I was talking about:

$ define /TRANSLATION_ATTRIBUTES=TERMINAL

The CONCEALED attribute is a different matter - it is for display purposes, but, yes, the combination of both attributes has always been 'interesting'.

VAX/VMS V3 had 3 fixed logical name tables (process?, group and system). One could not see any other tables.

[Jan van den Ende]

Uwe,


VAX/VMS V3 had 3 fixed logical name tables (process?, group and system).

make that four:
Process, Job, Group, and System; only those, and searched in that fixed order.


The CONCEALED attribute is a different matter - it is for display purposes, but, yes, the combination of both attributes has always been 'interesting'.


well... "display" purposes... not untrue, but it mostly serves to let directories be presented as devices, allowing for many "logical", and dynamic, devices mapped upon one physical.
Your interpunction of 'interesting' suggest things like puzzling, confusing, troublesome, or the likes of that.
Especially after the obligation for /TERMINAL has been lifted from /TRANS=CONCEAL, I hardly found any issue, and I grew really fond of the concept, even up to the degree that I count it as one of the things that makes VMS so great and so flexible.
(only one wish still standing: recursiveness, ie, allowing a directory within a concealed device to be the root of another one)


Proost.

Have one on me.

jpe

[Steve Nimr]

Be aware some 4100's get a second round of the same disks when using f$device("*","disk")

[Ian Miller.]

Steve, you may see the same disk names more than once from f$device if the alpha is connected to the disk and the disk is also served by another alpha which is also connected to the disk.

[Jan van den Ende]

Steve, Ian,

I have never noticed that myself, but I definitely would consider it a BUG.
If there are multiple paths to the same disk, then VMS should certainly be aware of that same-ness!!
I KNOW in our system there USED to be multiple paths from 2100's to HSZ40 connected disks, and THEY were seen only once.
Now (ES'es, DS) with HSG80 connected disks, each disk shows 5 (FIVE) paths to it (2 paths to both switches, plus an MSCP path), but each DISK is seen only once!

Proost.

Have one on me.

Seasonal greetings to all!

Jan

[Uwe Zessin]

It's an old problem since VAX/VMS version 4 that has to do with timing.