Hi,

for my project a need the information, what file operation (i.e. crea/dir or remove file) is been issued from an interactive terminal process.
I've seen that the "undocumented" dcl-command "set watch" just broadcasts this information to the terminal I need. I did look into the source and found that this program only modifies a system parameter called PIO$GW_DFPROT and CLASS_FLAGS[x], where x = XQP$_MAJOR_FUNC, XQP$_CNTRL_FUNC, XQP$_ATTRIBUTE,XQP$_DIR_OPS, XQP$_DUMP, XQP$_QUOTA or XQP$_PROTECTION are accessed.
It is unclear for me if it is possible to get access to these infos in an own written program. So I'm thankful for any ideas where to look at.

Eberhard

On 9/22/2023 3:08 AM, Eberhard Heuser wrote:
> for my project a need the information, what file operation (i.e.
> crea/dir or remove file) is been issued from an interactive terminal
> process. I've seen that the "undocumented" dcl-command "set watch"
> just broadcasts this information to the terminal I need. I did look
> into the source and found that this program only modifies a system
> parameter called PIO$GW_DFPROT and CLASS_FLAGS[x], where x =
> XQP$_MAJOR_FUNC, XQP$_CNTRL_FUNC, XQP$_ATTRIBUTE,XQP$_DIR_OPS,
> XQP$_DUMP, XQP$_QUOTA or XQP$_PROTECTION are accessed. It is unclear
> for me if it is possible to get access to these infos in an own
> written program. So I'm thankful for any ideas where to look at.

WIthout having seen any VMS source or know anything specific,
then it sounds like a program that:
* flip the bits that will triggers those unso messages to be sent
* trap all unso messages
would work.

The second part does not even require privs.

Arne

On 2023-09-22 07:08:56 +0000, Eberhard Heuser said:

> Hi,
>
> for my project a need the information, what file operation (i.e.
> crea/dir or remove file) is been issued from an interactive terminal
> process.
> I've seen that the "undocumented" dcl-command "set watch" just
> broadcasts this information to the terminal I need. I did look into the
> source and found that this program only modifies a system parameter
> called PIO$GW_DFPROT and CLASS_FLAGS[x], where x = XQP$_MAJOR_FUNC,
> XQP$_CNTRL_FUNC, XQP$_ATTRIBUTE,XQP$_DIR_OPS, XQP$_DUMP, XQP$_QUOTA or
> XQP$_PROTECTION are accessed.
> It is unclear for me if it is possible to get access to these infos in
> an own written program. So I'm thankful for any ideas where to look at.
>
> Eberhard

Sounds like something interesting I need to hack!

Brian Schenkenberger schrieb am Sonntag, 24. September 2023 um 22:36:09 UTC+2:
>> Sounds like something interesting I need to hack!

If you're willing to produce some code, that will be great!

My problem is that I cannot find an hook, that helps me to get the information, which
file operation (i.e. create/remove file / create/remove directory / copy file)
been isssued via a DCL-command. The drivers only care about I/O of records so
my hook must lay between the command layer and the driver I/O. I assume that F11ACP
is the part that might give this info back.

Eberhard

On Monday, September 25, 2023 at 9:26:39 AM UTC-4, Eberhard Heuser wrote:
> Brian Schenkenberger schrieb am Sonntag, 24. September 2023 um 22:36:09 UTC+2:
> >> Sounds like something interesting I need to hack!
> If you're willing to produce some code, that will be great!
>
> My problem is that I cannot find an hook, that helps me to get the information, which
> file operation (i.e. create/remove file / create/remove directory / copy file)
> been isssued via a DCL-command. The drivers only care about I/O of records so
> my hook must lay between the command layer and the driver I/O. I assume that F11ACP
> is the part that might give this info back.
>
> Eberhard
Here's the code to turn on the bits. The code in the XQP writes to SYS$OUTPUT so you should be able to
easily redirect it to a file, etc.

/*
$ search XGI5_RESD$:[F11X.SRC]FCPDEF.REQ xqp$_

XQP$_MAJOR_FUNC = 0, ! Notify about major events
XQP$_CNTRL_FUNC = 1, ! Notify about control functions
XQP$_ATTRIBUTE = 2, ! Notify about attribute modifications
XQP$_DIR_OPS = 3, ! Notify about directory operations
XQP$_DUMP = 4, ! Notify about misc things
XQP$_QUOTA = 5, ! Notify about quota functions
XQP$_PROTECTION = 6; ! Notify about protection checks
*/

extern short PIO$GW_DFPROT;
#include <stdio.h>
#include <starlet.h>

int flip() {
PIO$GW_DFPROT = PIO$GW_DFPROT | 1; // turn on XQP$_MAJOR_FUNC
return 1;
}

int main() {
int s;
printf("DFPROT = %hx\n",PIO$GW_DFPROT);
s = SYS$CMKRNL(&flip,0);
printf("CMKRNL returned %08x\n",s);
printf("DFPROT = %hx\n",PIO$GW_DFPROT);
}

John Reagan schrieb am Dienstag, 26. September 2023 um 03:02:50 UTC+2:
> On Monday, September 25, 2023 at 9:26:39 AM UTC-4, Eberhard Heuser wrote:
> > Brian Schenkenberger schrieb am Sonntag, 24. September 2023 um 22:36:09 UTC+2:
> > >> Sounds like something interesting I need to hack!
> > If you're willing to produce some code, that will be great!
> >
> > My problem is that I cannot find an hook, that helps me to get the information, which
> > file operation (i.e. create/remove file / create/remove directory / copy file)
> > been isssued via a DCL-command. The drivers only care about I/O of records so
> > my hook must lay between the command layer and the driver I/O. I assume that F11ACP
> > is the part that might give this info back.
> >
> > Eberhard
> Here's the code to turn on the bits. The code in the XQP writes to SYS$OUTPUT so you should be able to
> easily redirect it to a file, etc.
>
> /*
> $ search XGI5_RESD$:[F11X.SRC]FCPDEF.REQ xqp$_
>
> XQP$_MAJOR_FUNC = 0, ! Notify about major events
> XQP$_CNTRL_FUNC = 1, ! Notify about control functions
> XQP$_ATTRIBUTE = 2, ! Notify about attribute modifications
> XQP$_DIR_OPS = 3, ! Notify about directory operations
> XQP$_DUMP = 4, ! Notify about misc things
> XQP$_QUOTA = 5, ! Notify about quota functions
> XQP$_PROTECTION = 6; ! Notify about protection checks
> */
>
> extern short PIO$GW_DFPROT;
> #include <stdio.h>
> #include <starlet.h>
>
> int flip() {
> PIO$GW_DFPROT = PIO$GW_DFPROT | 1; // turn on XQP$_MAJOR_FUNC
> return 1;
> }
>
> int main() {
> int s;
> printf("DFPROT = %hx\n",PIO$GW_DFPROT);
> s = SYS$CMKRNL(&flip,0);
> printf("CMKRNL returned %08x\n",s);
> printf("DFPROT = %hx\n",PIO$GW_DFPROT);
> }

I'm sorry, but the redirection from sys$output has no effect. The XQP output
still appears on the terminal screen. I've tried several options

define/sys sys$output t.dat
define/sys/exe sys$output t.dat
define/proc sys$output t.dat

The same behavior with sys$error

So the question is what output channel is been used by QXP?

Eberhard

Eberhard Heuser schrieb am Dienstag, 26. September 2023 um 16:01:15 UTC+2:

> So the question is what output channel is been used by QXP?

Eberhard,

the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a supervisor mode AST) writes the output to SYS$OUTPUT using a RMS $PUT call.

By default SYS$OUTPUT is a supervisor mode confined logical name, which points to the 'terminal device'.

Question: can this be re-directed ?

Volker.

Yes, "define/super" redirects the output to a file.

Am 26.09.2023 um 17:53 schrieb Volker Halle via Info-vax:
> Eberhard Heuser schrieb am Dienstag, 26. September 2023 um 16:01:15 UTC+2:
>
>> So the question is what output channel is been used by QXP?
> Eberhard,
>
> the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a supervisor mode AST) writes the output to SYS$OUTPUT using a RMS $PUT call.
>
> By default SYS$OUTPUT is a supervisor mode confined logical name, which points to the 'terminal device'.
>
> Question: can this be re-directed ?
>
> Volker.
> _______________________________________________
> Info-vax mailing list
> Info-vax@rbnsn.com
> http://rbnsn.com/mailman/listinfo/info-vax_rbnsn.com
>

On 26/09/2023 18:28, vaxinf@gmail.com wrote:
> Yes, "define/super" redirects the output to a file.
>
> Am 26.09.2023 um 17:53 schrieb Volker Halle via Info-vax:
>> Eberhard Heuser schrieb am Dienstag, 26. September 2023 um 16:01:15
>> UTC+2:
>>
>>> So the question is what output channel is been used by QXP?
>> Eberhard,
>>
>> the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a
>> supervisor mode AST) writes the output to SYS$OUTPUT using a RMS $PUT
>> call.
>>
>> By default SYS$OUTPUT is a supervisor mode confined logical name,
>> which points to the 'terminal device'.
>>
>> Question: can this be re-directed ?
>>
>> Volker.
>> _______________________________________________
>> Info-vax mailing list
>> Info-vax@rbnsn.com
>> http://rbnsn.com/mailman/listinfo/info-vax_rbnsn.com
>>
>
>
I thought that was the default

--
Chris

On 9/26/2023 11:53 AM, Volker Halle wrote:
> Eberhard Heuser schrieb am Dienstag, 26. September 2023 um 16:01:15
> UTC+2:
>> So the question is what output channel is been used by QXP?
>
> the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a
> supervisor mode AST) writes the output to SYS$OUTPUT using a RMS $PUT
> call.
>
> By default SYS$OUTPUT is a supervisor mode confined logical name,
> which points to the 'terminal device'.
>
> Question: can this be re-directed ?

Another question: is SET WATCH smart enough not to notify about
disk IO triggered by its own notifications if they are redirected
to disk file.

Arne

On Tuesday, September 26, 2023 at 6:30:06 PM UTC+1, vax...@googlemail.com wrote:
> Yes, "define/super" redirects the output to a file.
> Am 26.09.2023 um 17:53 schrieb Volker Halle via Info-vax:
> > Eberhard Heuser schrieb am Dienstag, 26. September 2023 um 16:01:15 UTC+2:
> >
> >> So the question is what output channel is been used by QXP?
> > Eberhard,
> >
> > the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a supervisor mode AST) writes the output to SYS$OUTPUT using a RMS $PUT call.
> >
> > By default SYS$OUTPUT is a supervisor mode confined logical name, which points to the 'terminal device'.
> >
> > Question: can this be re-directed ?
> >
> > Volker.
> > _______________________________________________
> > Info-vax mailing list
> > Info...@rbnsn.com
> > http://rbnsn.com/mailman/listinfo/info-vax_rbnsn.com
> >

You can create a mailbox, define SYS$OUTPUT to that mailbox then read the mailbox to see the file watch messages from the XQP. Of course you see everything that is sent to SYS$OUTPUT but the XQP File Watch messages all start with %XQP so can be detected.

On Tuesday, September 26, 2023 at 11:43:49 PM UTC+1, Arne Vajhøj wrote:
> On 9/26/2023 11:53 AM, Volker Halle wrote:
> > Eberhard Heuser schrieb am Dienstag, 26. September 2023 um 16:01:15
> > UTC+2:
> >> So the question is what output channel is been used by QXP?
> >
> > the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a
> > supervisor mode AST) writes the output to SYS$OUTPUT using a RMS $PUT
> > call.
> >
> > By default SYS$OUTPUT is a supervisor mode confined logical name,
> > which points to the 'terminal device'.
> >
> > Question: can this be re-directed ?
> Another question: is SET WATCH smart enough not to notify about
> disk IO triggered by its own notifications if they are redirected
> to disk file.
>
> Arne

I expect not

Here is an example:

AXPVMS $ define/super sys$output xqp.txt
AXPVMS $ set watch file/class=major
AXPVMS $ dir xqp.txt
AXPVMS $ set watch file/class=nomajor
AXPVMS $ deass/super sys$output
AXPVMS $ ty xqp.txt

%XQP, Thread #0, Access (0,0,0) Status: 00000910
%XQP, Thread #0, Access SETWATCH.EXE;1 (1110,3,0) Status: 00000001
%XQP, Thread #0, Control function (1110,3,0) Status: 00000001
%XQP, Thread #0, Modify XQP.TXT;1 (39415,72,0) Status: 00000001
%XQP, Thread #0, Deaccess (1110,3,0) Reads: 6, Writes: 0, Status: 00000001
%XQP, Thread #0, Control function (39415,72,0) Status: 00000001
%XQP, Thread #0, Access HALLE.DIR;1 (2955,45459,0) Status: 00000001
%XQP, Thread #0, Final status: A6000001
%XQP, Thread #0, Deaccess (2955,45459,0) Reads: 1, Writes: 0, Status: 00000001

Directory USERDISK1:[HALLE]

XQP.TXT;1

Total of 1 file.
%XQP, Thread #0, Access (0,0,0) Status: 00000910
%XQP, Thread #0, Access SETWATCH.EXE;1 (1110,3,0) Status: 00000001
%XQP, Thread #0, Control function (1110,3,0) Status: 00000001

Volker.

On 9/27/2023 6:45 AM, Ian Miller wrote:
> On Tuesday, September 26, 2023 at 6:30:06 PM UTC+1,
> vax...@googlemail.com wrote:
>> Am 26.09.2023 um 17:53 schrieb Volker Halle via Info-vax:
>>> Eberhard Heuser schrieb am Dienstag, 26. September 2023 um
>>> 16:01:15 UTC+2:
>>>> So the question is what output channel is been used by QXP?
>>>
>>> the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a
>>> supervisor mode AST) writes the output to SYS$OUTPUT using a RMS
>>> $PUT call.
>>>
>>> By default SYS$OUTPUT is a supervisor mode confined logical name,
>>> which points to the 'terminal device'.
>>>
>>> Question: can this be re-directed ?
>>
>> Yes, "define/super" redirects the output to a file.
>
> You can create a mailbox, define SYS$OUTPUT to that mailbox then read
> the mailbox to see the file watch messages from the XQP. Of course
> you see everything that is sent to SYS$OUTPUT but the XQP File Watch
> messages all start with %XQP so can be detected.

Yes. But it looks like the define need to be done by DCL.

I did a little test trying to create the mailbox and defining
the logical from code and I don't think XQP does a SYS$OPEN
on whatever SYS$OUTPUT points to - I think it uses the
the first 4 bytes of the SYS$OUTPUT logical value to find a
reference to the file DCL already opened and just
SYS$PUT to that.

And I don't know how to stuff a reference to a newly
created mailbox in there programmatically.

Arne

On 9/30/2023 10:03 AM, Arne Vajhøj wrote:
> On 9/27/2023 6:45 AM, Ian Miller wrote:
>> On Tuesday, September 26, 2023 at 6:30:06 PM UTC+1,
>> vax...@googlemail.com wrote:
>>> Am 26.09.2023 um 17:53 schrieb Volker Halle via Info-vax:
>>>> Eberhard Heuser schrieb am Dienstag, 26. September 2023 um
>>>> 16:01:15 UTC+2:
>>>>> So the question is what output channel is been used by QXP?
>>>>
>>>> the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a
>>>> supervisor mode AST) writes the output to SYS$OUTPUT using a RMS
>>>> $PUT call.
>>>>
>>>> By default SYS$OUTPUT is a supervisor mode confined logical name,
>>>> which points to the 'terminal device'.
>>>>
>>>> Question: can this be re-directed ?
> >>
> >> Yes, "define/super" redirects the output to a file.
>>
>> You can create a mailbox, define SYS$OUTPUT to that mailbox then read
>> the mailbox to see the file watch messages from the XQP.  Of course
>> you see everything that is sent to SYS$OUTPUT but the XQP File Watch
>> messages all start with %XQP so can be detected.
>
> Yes. But it looks like the define need to be done by DCL.
>
> I did a little test trying to create the mailbox and defining
> the logical from code and I don't think XQP does a SYS$OPEN
> on whatever SYS$OUTPUT points to - I think it uses the
> the first 4 bytes of the SYS$OUTPUT logical value to find a
> reference to the file DCL already opened and just
> SYS$PUT to that.
>
> And I don't know how to stuff a reference to a newly
> created mailbox in there programmatically.

Or to rephrase - I can not find:
LIB$OPEN_PROCESS_PERMANENT_FILE

:-)

Arne

On 30/09/2023 16:32, Arne Vajhøj wrote:
> On 9/30/2023 10:03 AM, Arne Vajhøj wrote:
>> On 9/27/2023 6:45 AM, Ian Miller wrote:
>>> On Tuesday, September 26, 2023 at 6:30:06 PM UTC+1,
>>> vax...@googlemail.com wrote:
>>>> Am 26.09.2023 um 17:53 schrieb Volker Halle via Info-vax:
>>>>> Eberhard Heuser schrieb am Dienstag, 26. September 2023 um
>>>>> 16:01:15 UTC+2:
>>>>>> So the question is what output channel is been used by QXP?
>>>>>
>>>>> the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a
>>>>> supervisor mode AST) writes the output to SYS$OUTPUT using a RMS
>>>>> $PUT call.
>>>>>
>>>>> By default SYS$OUTPUT is a supervisor mode confined logical name,
>>>>> which points to the 'terminal device'.
>>>>>
>>>>> Question: can this be re-directed ?
>>  >>
>>  >> Yes, "define/super" redirects the output to a file.
>>>
>>> You can create a mailbox, define SYS$OUTPUT to that mailbox then read
>>> the mailbox to see the file watch messages from the XQP.  Of course
>>> you see everything that is sent to SYS$OUTPUT but the XQP File Watch
>>> messages all start with %XQP so can be detected.
>>
>> Yes. But it looks like the define need to be done by DCL.
>>
>> I did a little test trying to create the mailbox and defining
>> the logical from code and I don't think XQP does a SYS$OPEN
>> on whatever SYS$OUTPUT points to - I think it uses the
>> the first 4 bytes of the SYS$OUTPUT logical value to find a
>> reference to the file DCL already opened and just
>> SYS$PUT to that.
>>
>> And I don't know how to stuff a reference to a newly
>> created mailbox in there programmatically.
>
> Or to rephrase - I can not find:
>    LIB$OPEN_PROCESS_PERMANENT_FILE
>
> :-)
>
> Arne
>

You could presumably set the logical with LIB$SET_LOGICAL

--
Chris

On 9/30/2023 1:34 PM, Chris Townley wrote:
> On 30/09/2023 16:32, Arne Vajhøj wrote:
>> On 9/30/2023 10:03 AM, Arne Vajhøj wrote:
>>> On 9/27/2023 6:45 AM, Ian Miller wrote:
>>>> On Tuesday, September 26, 2023 at 6:30:06 PM UTC+1,
>>>> vax...@googlemail.com wrote:
>>>>> Am 26.09.2023 um 17:53 schrieb Volker Halle via Info-vax:
>>>>>> Eberhard Heuser schrieb am Dienstag, 26. September 2023 um
>>>>>> 16:01:15 UTC+2:
>>>>>>> So the question is what output channel is been used by QXP?
>>>>>>
>>>>>> the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a
>>>>>> supervisor mode AST) writes the output to SYS$OUTPUT using a RMS
>>>>>> $PUT call.
>>>>>>
>>>>>> By default SYS$OUTPUT is a supervisor mode confined logical name,
>>>>>> which points to the 'terminal device'.
>>>>>>
>>>>>> Question: can this be re-directed ?
>>>  >>
>>>  >> Yes, "define/super" redirects the output to a file.
>>>>
>>>> You can create a mailbox, define SYS$OUTPUT to that mailbox then read
>>>> the mailbox to see the file watch messages from the XQP.  Of course
>>>> you see everything that is sent to SYS$OUTPUT but the XQP File Watch
>>>> messages all start with %XQP so can be detected.
>>>
>>> Yes. But it looks like the define need to be done by DCL.
>>>
>>> I did a little test trying to create the mailbox and defining
>>> the logical from code and I don't think XQP does a SYS$OPEN
>>> on whatever SYS$OUTPUT points to - I think it uses the
>>> the first 4 bytes of the SYS$OUTPUT logical value to find a
>>> reference to the file DCL already opened and just
>>> SYS$PUT to that.
>>>
>>> And I don't know how to stuff a reference to a newly
>>> created mailbox in there programmatically.
>>
>> Or to rephrase - I can not find:
>>     LIB$OPEN_PROCESS_PERMANENT_FILE
>>
>> :-)
>
> You could presumably set the logical with LIB$SET_LOGICAL

That is not the problem.

The problem is to get a valid value to set.

Arne

On Saturday, September 30, 2023 at 6:34:25 PM UTC+1, Chris Townley wrote:
> On 30/09/2023 16:32, Arne Vajhøj wrote:
> > On 9/30/2023 10:03 AM, Arne Vajhøj wrote:
> >> On 9/27/2023 6:45 AM, Ian Miller wrote:
> >>> On Tuesday, September 26, 2023 at 6:30:06 PM UTC+1,
> >>> vax...@googlemail.com wrote:
> >>>> Am 26.09.2023 um 17:53 schrieb Volker Halle via Info-vax:
> >>>>> Eberhard Heuser schrieb am Dienstag, 26. September 2023 um
> >>>>> 16:01:15 UTC+2:
> >>>>>> So the question is what output channel is been used by QXP?
> >>>>>
> >>>>> the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a
> >>>>> supervisor mode AST) writes the output to SYS$OUTPUT using a RMS
> >>>>> $PUT call.
> >>>>>
> >>>>> By default SYS$OUTPUT is a supervisor mode confined logical name,
> >>>>> which points to the 'terminal device'.
> >>>>>
> >>>>> Question: can this be re-directed ?
> >> >>
> >> >> Yes, "define/super" redirects the output to a file.
> >>>
> >>> You can create a mailbox, define SYS$OUTPUT to that mailbox then read
> >>> the mailbox to see the file watch messages from the XQP. Of course
> >>> you see everything that is sent to SYS$OUTPUT but the XQP File Watch
> >>> messages all start with %XQP so can be detected.
> >>
> >> Yes. But it looks like the define need to be done by DCL.
> >>
> >> I did a little test trying to create the mailbox and defining
> >> the logical from code and I don't think XQP does a SYS$OPEN
> >> on whatever SYS$OUTPUT points to - I think it uses the
> >> the first 4 bytes of the SYS$OUTPUT logical value to find a
> >> reference to the file DCL already opened and just
> >> SYS$PUT to that.
> >>
> >> And I don't know how to stuff a reference to a newly
> >> created mailbox in there programmatically.
> >
> > Or to rephrase - I can not find:
> > LIB$OPEN_PROCESS_PERMANENT_FILE
> >
> > :-)
> >
> > Arne
> >
> You could presumably set the logical with LIB$SET_LOGICAL
>
> --
> Chris

yes in Super/exec/kernel mode specifying the mode of the logical name as Supervisor.

On Saturday, September 30, 2023 at 3:03:11 PM UTC+1, Arne Vajhøj wrote:
> On 9/27/2023 6:45 AM, Ian Miller wrote:
> > On Tuesday, September 26, 2023 at 6:30:06 PM UTC+1,
> > vax...@googlemail.com wrote:
> >> Am 26.09.2023 um 17:53 schrieb Volker Halle via Info-vax:
> >>> Eberhard Heuser schrieb am Dienstag, 26. September 2023 um
> >>> 16:01:15 UTC+2:
> >>>> So the question is what output channel is been used by QXP?
> >>>
> >>> the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a
> >>> supervisor mode AST) writes the output to SYS$OUTPUT using a RMS
> >>> $PUT call.
> >>>
> >>> By default SYS$OUTPUT is a supervisor mode confined logical name,
> >>> which points to the 'terminal device'.
> >>>
> >>> Question: can this be re-directed ?
> >>
> >> Yes, "define/super" redirects the output to a file.
> >
> > You can create a mailbox, define SYS$OUTPUT to that mailbox then read
> > the mailbox to see the file watch messages from the XQP. Of course
> > you see everything that is sent to SYS$OUTPUT but the XQP File Watch
> > messages all start with %XQP so can be detected.
> Yes. But it looks like the define need to be done by DCL.
>
> I did a little test trying to create the mailbox and defining
> the logical from code and I don't think XQP does a SYS$OPEN
> on whatever SYS$OUTPUT points to - I think it uses the
> the first 4 bytes of the SYS$OUTPUT logical value to find a
> reference to the file DCL already opened and just
> SYS$PUT to that.
>
> And I don't know how to stuff a reference to a newly
> created mailbox in there programmatically.
>
> Arne

Call LNM$CRELNM in Super/exec/kernel mode specifying the mode of the logical name as Supervisor which would redefine SYS$OUTPUT to your mailbox.

On 10/2/2023 9:48 AM, Ian Miller wrote:
> On Saturday, September 30, 2023 at 3:03:11 PM UTC+1, Arne Vajhøj wrote:
>> On 9/27/2023 6:45 AM, Ian Miller wrote:
>>> On Tuesday, September 26, 2023 at 6:30:06 PM UTC+1,
>>> vax...@googlemail.com wrote:
>>>> Am 26.09.2023 um 17:53 schrieb Volker Halle via Info-vax:
>>>>> Eberhard Heuser schrieb am Dienstag, 26. September 2023 um
>>>>> 16:01:15 UTC+2:
>>>>>> So the question is what output channel is been used by QXP?
>>>>>
>>>>> the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a
>>>>> supervisor mode AST) writes the output to SYS$OUTPUT using a RMS
>>>>> $PUT call.
>>>>>
>>>>> By default SYS$OUTPUT is a supervisor mode confined logical name,
>>>>> which points to the 'terminal device'.
>>>>>
>>>>> Question: can this be re-directed ?
>>>>
>>>> Yes, "define/super" redirects the output to a file.
>>>
>>> You can create a mailbox, define SYS$OUTPUT to that mailbox then read
>>> the mailbox to see the file watch messages from the XQP. Of course
>>> you see everything that is sent to SYS$OUTPUT but the XQP File Watch
>>> messages all start with %XQP so can be detected.
>> Yes. But it looks like the define need to be done by DCL.
>>
>> I did a little test trying to create the mailbox and defining
>> the logical from code and I don't think XQP does a SYS$OPEN
>> on whatever SYS$OUTPUT points to - I think it uses the
>> the first 4 bytes of the SYS$OUTPUT logical value to find a
>> reference to the file DCL already opened and just
>> SYS$PUT to that.
>>
>> And I don't know how to stuff a reference to a newly
>> created mailbox in there programmatically.
>
> Call LNM$CRELNM in Super/exec/kernel mode specifying the mode of the
> logical name as Supervisor which would redefine SYS$OUTPUT to your
> mailbox.
The problem is not defining SYS$OUTPUT in supervisor
mode. That is easy.

But I cannot get it to use the value I define.

Either I am doing it wrong or that XQP code is doing
something funky.

I am suspecting that it uses the IFI in the logical
to find FAB & RAB in P1 space and use those instead of
just calling SYS$OPEN/SYS$CREATE and SYS$CONNECT
to setup new FAB & RAB.

Arne

On 10/2/2023 4:10 PM, Arne Vajhøj wrote:
> On 10/2/2023 9:48 AM, Ian Miller wrote:
>> On Saturday, September 30, 2023 at 3:03:11 PM UTC+1, Arne Vajhøj wrote:
>>> On 9/27/2023 6:45 AM, Ian Miller wrote:
>>>> On Tuesday, September 26, 2023 at 6:30:06 PM UTC+1,
>>>> vax...@googlemail.com wrote:
>>>>> Am 26.09.2023 um 17:53 schrieb Volker Halle via Info-vax:
>>>>>> Eberhard Heuser schrieb am Dienstag, 26. September 2023 um
>>>>>> 16:01:15 UTC+2:
>>>>>>> So the question is what output channel is been used by QXP?
>>>>>>
>>>>>> the XQP (in module [F11X]DISPAT) in routine NOTIFY_AST (run as a
>>>>>> supervisor mode AST) writes the output to SYS$OUTPUT using a RMS
>>>>>> $PUT call.
>>>>>>
>>>>>> By default SYS$OUTPUT is a supervisor mode confined logical name,
>>>>>> which points to the 'terminal device'.
>>>>>>
>>>>>> Question: can this be re-directed ?
>>>>>
>>>>> Yes, "define/super" redirects the output to a file.
>>>>
>>>> You can create a mailbox, define SYS$OUTPUT to that mailbox then read
>>>> the mailbox to see the file watch messages from the XQP. Of course
>>>> you see everything that is sent to SYS$OUTPUT but the XQP File Watch
>>>> messages all start with %XQP so can be detected.
>>> Yes. But it looks like the define need to be done by DCL.
>>>
>>> I did a little test trying to create the mailbox and defining
>>> the logical from code and I don't think XQP does a SYS$OPEN
>>> on whatever SYS$OUTPUT points to - I think it uses the
>>> the first 4 bytes of the SYS$OUTPUT logical value to find a
>>> reference to the file DCL already opened and just
>>> SYS$PUT to that.
>>>
>>> And I don't know how to stuff a reference to a newly
>>> created mailbox in there programmatically.
>>
>> Call LNM$CRELNM in Super/exec/kernel mode specifying the mode of the
>> logical name as Supervisor which would redefine SYS$OUTPUT to your
>> mailbox.
> The problem is not defining SYS$OUTPUT in supervisor
> mode. That is easy.
>
> But I cannot get it to use the value I define.
>
> Either I am doing it wrong or that XQP code is doing
> something funky.
>
> I am suspecting that it uses the IFI in the logical
> to find FAB & RAB in P1 space and use those instead of
> just calling SYS$OPEN/SYS$CREATE and SYS$CONNECT
> to setup new FAB & RAB.

Or maybe something else.

I create a mailbox and define SYS$OUTPUT to
point to that, use Reagans code to flip bit
and what I observe is that:
- plain C IO end up fine in the mailbox
- if SET WATCH has not been used before by process
then the process hangs
- if SET WATCH has been used before with terminal
then the output goes to terminal ignoring the
define

Weird.

Arne

Arne Vajhøj schrieb am Dienstag, 3. Oktober 2023 um 01:48:53 UTC+2:
> >
> > I am suspecting that it uses the IFI in the logical
> > to find FAB & RAB in P1 space and use those instead of
> > just calling SYS$OPEN/SYS$CREATE and SYS$CONNECT
> > to setup new FAB & RAB.
>
> I create a mailbox and define SYS$OUTPUT to
> point to that, use Reagans code to flip bit
> and what I observe is that:
> - plain C IO end up fine in the mailbox
> - if SET WATCH has not been used before by process
> then the process hangs
> - if SET WATCH has been used before with terminal
> then the output goes to terminal ignoring the
> define

Arne,

the NOTIFY_AST checks, if it's output file has already been opened (by checking it's internal IFI/ISI storage location) and opens the file pointed to by SYS$OUTPUT, if it has not already been opened by the XQP. It then remembers the IFI/ISI for the next invocation of the AST.

Do you have some code set up to READ the mailbox ? Otherwise the process trying to write to a full mailbox will hang...

Volker.

On 10/3/2023 3:31 AM, Volker Halle wrote:
> Arne Vajhøj schrieb am Dienstag, 3. Oktober 2023 um 01:48:53 UTC+2:
>>>
>>> I am suspecting that it uses the IFI in the logical to find FAB &
>>> RAB in P1 space and use those instead of just calling
>>> SYS$OPEN/SYS$CREATE and SYS$CONNECT to setup new FAB & RAB.
>>
>> I create a mailbox and define SYS$OUTPUT to point to that, use
>> Reagans code to flip bit and what I observe is that: - plain C IO
>> end up fine in the mailbox - if SET WATCH has not been used before
>> by process then the process hangs - if SET WATCH has been used
>> before with terminal then the output goes to terminal ignoring the
>> define

> the NOTIFY_AST checks, if it's output file has already been opened
> (by checking it's internal IFI/ISI storage location) and opens the
> file pointed to by SYS$OUTPUT, if it has not already been opened by
> the XQP. It then remembers the IFI/ISI for the next invocation of the
> AST.
>
> Do you have some code set up to READ the mailbox ? Otherwise the
> process trying to write to a full mailbox will hang...

Yes - I have a write attention AST doing reads.

But it sounds like there must be a problem in my code.

I will have to look at again, clean it up and if I am lost
post an illustrative example.

Arne