Just being playing with a routine to squirrel away the command buffer on
logout, which be quite useful after a system maintenance session.

However when run in a command procedure, it just doesn't produce anything.

Is this normal?

BTW on E9.2-1

--
Chris

On 22/06/2023 18:35, Chris Townley wrote:
> Just being playing with a routine to squirrel away the command buffer on
> logout, which be quite useful after a system maintenance session.
>
> However when run in a command procedure, it just doesn't produce anything.
>
> Is this normal?
>
> BTW on E9.2-1
>

Just checked on AXP V8.4-2L1 and that is the same. Using the /output
qualifier.

I could understand it not being available with other than the output
qualifier, or perhaps /ERASE, but seems a bit off!

--
Chris

On Thursday, June 22, 2023 at 1:47:37 PM UTC-4, Chris Townley wrote:
> On 22/06/2023 18:35, Chris Townley wrote:
> > Just being playing with a routine to squirrel away the command buffer on
> > logout, which be quite useful after a system maintenance session.
> >
> > However when run in a command procedure, it just doesn't produce anything.
> >
> > Is this normal?
> >
> > BTW on E9.2-1
> >
> Just checked on AXP V8.4-2L1 and that is the same. Using the /output
> qualifier.
>
> I could understand it not being available with other than the output
> qualifier, or perhaps /ERASE, but seems a bit off!
>
> --
> Chris

RECALL demands that the caller be at procedure depth 0 (the DCL prompt).
Execution within a command procedure has a procedure depth of 1 (or
more if a command procedure call another, etc). See

$ write sys$output f$environment("depth")

See previous similar discussion in

https://community.hpe.com/t5/operating-system-openvms/recall-command/td-p/4947602

Volker.

On Thursday, June 22, 2023 at 2:40:03 PM UTC-4, Volker Halle wrote:
> See previous similar discussion in
>
> https://community.hpe.com/t5/operating-system-openvms/recall-command/td-p/4947602
>
> Volker.

And https://groups.google.com/g/comp.os.vms/c/a1SlLIzkQm0/m/aV4-sHxgAgAJ
might also be of interest.

On 22/06/2023 19:40, Volker Halle wrote:
> See previous similar discussion in
>
> https://community.hpe.com/t5/operating-system-openvms/recall-command/td-p/4947602
>
> Volker.

Thanks, that gives some perspective, and possible workarounds. I have in
the past used set host/log but as said, you need to remember!

--
Chris

You can define a logout symbol:
"logout:==pipe recall/out=session.history ; eoj"
or a different symbol :
"bye:==pipe recall/out=session.history ; logout/full".

On 6/22/2023 2:06 PM, Jim wrote:
> On Thursday, June 22, 2023 at 1:47:37 PM UTC-4, Chris Townley wrote:
>> On 22/06/2023 18:35, Chris Townley wrote:
>>> Just being playing with a routine to squirrel away the command buffer on
>>> logout, which be quite useful after a system maintenance session.
>>>
>>> However when run in a command procedure, it just doesn't produce anything.
>>>
>>> Is this normal?
>>>
>>> BTW on E9.2-1
>>>
>> Just checked on AXP V8.4-2L1 and that is the same. Using the /output
>> qualifier.
>>
>> I could understand it not being available with other than the output
>> qualifier, or perhaps /ERASE, but seems a bit off!
>
> RECALL demands that the caller be at procedure depth 0 (the DCL prompt).
> Execution within a command procedure has a procedure depth of 1 (or
> more if a command procedure call another, etc). See
>
> $ write sys$output f$environment("depth")

DCL may not be cooperative.

But the data is there, so if one is willing to wander
into unsupported territory, then ...

This code seems to accidental work on VMS Alpha 8.4:

.title sys2
PRC_C_CMDBUFSIZ=4097
.psect $PDATA quad,pic,con,lcl,shr,noexe,nowrt
; buf1: .blkb PRC_C_CMDBUFSIZ ; first copy of recall-buffer
buf2: .blkb PRC_C_CMDBUFSIZ ; second copy of recall-buffer
.psect $CODE quad,pic,con,lcl,shr,exe,nowrt
;***************************************
; ; SYS2_RECALL ( CMDSTR , CMDLEN , NCMD )
; ; recall commands
; ;***************************************
.entry sys2_recall,^m<r2,r3,r4,r5,r6,r7,r8,r9,r10,r11>
movl B^4(ap),r11
cvtwl (r11),r8 ; r8 = string length
movl B^4(r11),r11 ; r11 = string address
movl B^8(ap),r10 ; r10 = length address
movl #0,r9 ; r9 = count
movl #CTL$AG_CLIDATA,r0
movl W^PPD$L_PRC(r0),r0
movl W^PRC_L_RECALLPTR(r0),r7
moval W^PRC_G_COMMANDS(r0),r0
subl2 r0,r7
moval buf1,r1
movc3 #PRC_C_CMDBUFSIZ,(r0),(r1)
moval buf1,r1
moval buf2,r2
movc3 #PRC_C_CMDBUFSIZ,(r1),(r2)
moval buf2,r6
addl2 r6,r7 ; r7 = pointer in buf2
subl3 #PRC_C_CMDBUFSIZ,r7,r6 ; r6 = start in buf1
100$: subl2 #2,r7 ; move back to trailing length
cmpl r7,r6 ; if past start in buf1 done
blss 200$
tstw (r7) ; if zero length done
bleq 200$
cvtwl (r7),(r10) ; save length
subl2 (r10),r7
movc3 (r10),(r7),(r11) ; save string
addl2 #4,r10 ; increment length address
addl2 r8,r11 ; increment string address
incl r9 ; increment count
subl2 #4,r7 ; skip 4 bytes (new in later
VMS versions??)
brb 100$
200$: movl r9,@B^12(ap) ; save count
ret
.end

Test program:

program rdump
implicit none
integer*4 MAX_CMD
parameter (MAX_CMD=200)
character*256 cmd(MAX_CMD)
integer*4 lcmd(MAX_CMD), ncmd, i
call sys2_recall(cmd, lcmd, ncmd)
do 100 i = 1,ncmd
write(*,*) '|'//cmd(i)(1:lcmd(i))//'|'
100 continue
end

Arne

On 2023-06-22, hb@end.of.inter.net <h4711b@gmail.com> wrote:
> You can define a logout symbol:
> "logout:==pipe recall/out=session.history ; eoj"
> or a different symbol :
> "bye:==pipe recall/out=session.history ; logout/full".

The problem with these workarounds is what happens if you have more
than one session going at the same time and you need to preserve the
commands entered during all sessions ?

Simon.

--
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Walking destinations on a map are further away than they appear.

On 23/06/2023 13:09, Simon Clubley wrote:
> On 2023-06-22, hb@end.of.inter.net <h4711b@gmail.com> wrote:
>> You can define a logout symbol:
>> "logout:==pipe recall/out=session.history ; eoj"
>> or a different symbol :
>> "bye:==pipe recall/out=session.history ; logout/full".
>
> The problem with these workarounds is what happens if you have more
> than one session going at the same time and you need to preserve the
> commands entered during all sessions ?
>
> Simon.
>

VMS has a useful feature that allows for multiple file versions ;)

--
Chris

On 6/23/2023 9:09 AM, Chris Townley wrote:
> On 23/06/2023 13:09, Simon Clubley wrote:
>> On 2023-06-22, hb@end.of.inter.net <h4711b@gmail.com> wrote:
>>> You can define a logout symbol:
>>> "logout:==pipe recall/out=session.history ; eoj"
>>> or a different symbol :
>>> "bye:==pipe recall/out=session.history ; logout/full".
>>
>> The problem with these workarounds is what happens if you have more
>> than one session going at the same time and you need to preserve the
>> commands entered during all sessions ?
>>
>> Simon.
>>
>
> VMS has a useful feature that allows for multiple file versions ;)
>

And how do you tell which version is from which session?

bill

On 23/06/2023 14:14, bill wrote:
> On 6/23/2023 9:09 AM, Chris Townley wrote:
>> On 23/06/2023 13:09, Simon Clubley wrote:
>>> On 2023-06-22, hb@end.of.inter.net <h4711b@gmail.com> wrote:
>>>> You can define a logout symbol:
>>>> "logout:==pipe recall/out=session.history ; eoj"
>>>> or a different symbol :
>>>> "bye:==pipe recall/out=session.history ; logout/full".
>>>
>>> The problem with these workarounds is what happens if you have more
>>> than one session going at the same time and you need to preserve the
>>> commands entered during all sessions ?
>>>
>>> Simon.
>>>
>>
>> VMS has a useful feature that allows for multiple file versions ;)
>>
>
> And how do you tell which version is from which session?
>
> bill
>
Possibly with a bit of guesswork, or by looking at the contents.
Personally I want to be able to search for a command I previously used,
as I have done for years with *nix, so actually it doesn't really matter
which it was.

I am just checking out a way of naming the file for the login date
time, and the file creation date would give the logout time

--
Chris

On 2023-06-23, Chris Townley <news@cct-net.co.uk> wrote:
> Possibly with a bit of guesswork, or by looking at the contents.
> Personally I want to be able to search for a command I previously used,
> as I have done for years with *nix, so actually it doesn't really matter
> which it was.
>

With bash, it's much cleaner however, and I wish DCL had a bash-style
incremental search facility.

On Linux, I run with a command history buffer of 5,000 commands which
means all the regular commands never expire (both daily and across
multiple active projects), while allowing all the one-off junk to
eventually drop out of the command history buffer automatically.

One problem with your approach is that you will be saving all the
one-off junk as well as the commands you really want to save.

Simon.

--
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Walking destinations on a map are further away than they appear.

On 23/06/2023 18:28, Simon Clubley wrote:
> On 2023-06-23, Chris Townley <news@cct-net.co.uk> wrote:
>> Possibly with a bit of guesswork, or by looking at the contents.
>> Personally I want to be able to search for a command I previously used,
>> as I have done for years with *nix, so actually it doesn't really matter
>> which it was.
>>
>
> With bash, it's much cleaner however, and I wish DCL had a bash-style
> incremental search facility.
>
> On Linux, I run with a command history buffer of 5,000 commands which
> means all the regular commands never expire (both daily and across
> multiple active projects), while allowing all the one-off junk to
> eventually drop out of the command history buffer automatically.
>
> One problem with your approach is that you will be saving all the
> one-off junk as well as the commands you really want to save.
>
> Simon.
>

No problem with that. Just a load of files to GREP/SEARCH for that
stupidly complex that you worked out a few days/months ago!

If anybody is interested, I added the following to my LOGIN.COM for
interactive sessions:

$ if f$search("HIST.DIR") .eqs. "" then create/dire/log [.HIST]
$ hstnm = f$cvtime(f$getjpi("", "LOGINTIM"), "COMPARISON", "DATE") -
+ "_" + f$extract(0, 5, f$cvtime(f$getjpi("", -
"LOGINTIM"), "COMPARISON", "TIME")) - ":" + ".LIS"
$ hstnm = f$trnlnm("sys$login") - "]" + ".HIST]HST_''hstnm'"
$ lo*gout == "pipe recall/out=''histname' ; logouts /full "

On a home machine, could possibly add to SYLOGIN.COM
--
Chris

Den 2023-06-23 kl. 19:38, skrev Chris Townley:
> On 23/06/2023 18:28, Simon Clubley wrote:
>> On 2023-06-23, Chris Townley <news@cct-net.co.uk> wrote:
>>> Possibly with a bit of guesswork, or by looking at the contents.
>>> Personally I want to be able to search for a command I previously used,
>>> as I have done for years with *nix, so actually it doesn't really matter
>>> which it was.
>>>
>>
>> With bash, it's much cleaner however, and I wish DCL had a bash-style
>> incremental search facility.
>>
>> On Linux, I run with a command history buffer of 5,000 commands which
>> means all the regular commands never expire (both daily and across
>> multiple active projects), while allowing all the one-off junk to
>> eventually drop out of the command history buffer automatically.
>>
>> One problem with your approach is that you will be saving all the
>> one-off junk as well as the commands you really want to save.
>>
>> Simon.
>>
>
> No problem with that. Just a load of files to GREP/SEARCH for that
> stupidly complex that you worked out a few days/months ago!

Those commands should be in a COM file anyway.

On 2023-06-22 17:35:28 +0000, Chris Townley said:

> Just being playing with a routine to squirrel away the command buffer
> on logout, which be quite useful after a system maintenance session.
>
> However when run in a command procedure, it just doesn't produce anything.
>
> Is this normal?
>
> BTW on E9.2-1

Set the procedure level to 0 in your command procedure before executing
the RECALL commands.

File PROCLVL0.MAR is:

.TITLE PROCLVL0
.IDENT "V01.0 %2023.185"

$SSDEF

.ENTRY GO,0
$CMEXEC_S routin = PROCLVL0
RET

.ENTRY PROCLVL0,0

MOVAB @#CTL$AG_CLIDATA,R0 ; get the base of CLI data region
MOVL PPD$L_PRC(R0),R0 ; add address of CLI own storage
CLRW PRC_W_PROCLEVEL(R0) ; zap current DCL procedure level
MOVL #SS$_NORMAL,R0
RET

.END GO

$ MACRO PROCLVL0
$ LINK/SYSEXE PROCLVL0,SYS$LOADABLE_IMAGES:DCLDEF.STB

File LOGOUT_SAVE_RECALL.COM is:

$ RUN PROCLVL0
$ RECALL/ALL
$ RECALL/OUTPUT=RECALL.BUFFER
$ LOGOUT/FULL

Don't muck about in DCL too much after you've zapped its procedure level.

On Tuesday, July 4, 2023 at 11:14:46 AM UTC-4, Brian Schenkenberger wrote:
> On 2023-06-22 17:35:28 +0000, Chris Townley said:
>
> > Just being playing with a routine to squirrel away the command buffer
: > > However when run in a command procedure, it just doesn't produce anything.

> Set the procedure level to 0 in your command procedure before executing
: > CLRW PRC_W_PROCLEVEL(R0) ; zap current DCL procedure level
: > Don't muck about in DCL too much after you've zapped its procedure level.

Nice!
Hein.

On 7/4/2023 11:14 AM, Brian Schenkenberger wrote:
> On 2023-06-22 17:35:28 +0000, Chris Townley said:
>> Just being playing with a routine to squirrel away the command buffer
>> on logout, which be quite useful after a system maintenance session.
>>
>> However when run in a command procedure, it just doesn't produce
>> anything.

> Set the procedure level to 0 in your command procedure before executing
> the RECALL commands.
>
> File PROCLVL0.MAR is:
>
>     .TITLE    PROCLVL0
>     .IDENT    "V01.0 %2023.185"
>
>     $SSDEF
>
>     .ENTRY    GO,0
>     $CMEXEC_S    routin    = PROCLVL0
>     RET
>
>     .ENTRY    PROCLVL0,0
>
>     MOVAB    @#CTL$AG_CLIDATA,R0        ; get the base of CLI data region
>     MOVL    PPD$L_PRC(R0),R0        ; add address of CLI own storage
>     CLRW    PRC_W_PROCLEVEL(R0)    ; zap current DCL procedure level
>     MOVL    #SS$_NORMAL,R0
>     RET
>
>     .END    GO

Brilliant.

But note that privs are required.

Arne

On 2023-07-05 19:11:48 +0000, Arne Vajhj said:

> On 7/4/2023 11:14 AM, Brian Schenkenberger wrote:
>> On 2023-06-22 17:35:28 +0000, Chris Townley said:
>>> Just being playing with a routine to squirrel away the command buffer
>>> on logout, which be quite useful after a system maintenance session.
>>>
>>> However when run in a command procedure, it just doesn't produce anything.
>
>> Set the procedure level to 0 in your command procedure before executing
>> the RECALL commands.
>>
>> File PROCLVL0.MAR is:
>>
>>     .TITLE    PROCLVL0
>>     .IDENT    "V01.0 %2023.185"
>>
>>     $SSDEF
>>
>>     .ENTRY    GO,0
>>     $CMEXEC_S    routin    = PROCLVL0
>>     RET
>>
>>     .ENTRY    PROCLVL0,0
>>
>>     MOVAB    @#CTL$AG_CLIDATA,R0        ; get the base of CLI data region
>>     MOVL    PPD$L_PRC(R0),R0        ; add address of CLI own storage
>>     CLRW    PRC_W_PROCLEVEL(R0)    ; zap current DCL procedure level
>>     MOVL    #SS$_NORMAL,R0
>>     RET
>>
>>     .END    GO
>
> Brilliant.
>
> But note that privs are required.
>
> Arne

True! Most of DCL storage is URSW. Hopping to EXEC mode allows one to
change things that DCL would rather not have changed; at least, by the
user. ;)

This little snippet of code is rather inocuous; it can be installed
with EXEC privilege. Run from the command line, it doesn't do anything
because PROCLEVEL _is_ zero. However, when run in a command procedure,
exiting to the command line will logout the process. Thus, my
LOGOUT/FULL was really not necessary unless one wishes to see the
accounting report that it issues.