| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 866.1 | There is a memory leak, but not because titles aren't reused | HANNAH::MESSENGER | Bob Messenger | Thu Jun 01 1989 16:15 | 20 |
| Re: .0
If you quit out of DECterm 3, the DECterm controller deallocates all the
resources associated with that window. The number "3" has no significance;
it is not a slot number. The routine that sets the default title simply
increments a counter every time it's called. With a little extra work, it
could scan through the data structures for all the active windows and choose
the lowest available number, just so the user won't think he/she is being
cheated out of some resources. And no, there is no way to resurrect DECterm 3
once it's gone away (although the process might still be around as a detached
process, depending on how you've set things up).
There is one resource leak that I know of, though: DECterm doesn't free the
fonts that it loads, which eats up something like 128 pages for every three
DECterms you create. The only way to recover that memory is kill the DECterm
controller, which generally means quitting out of the session (although I think
if you just stop the controller process the session manager will restart it).
-- Bob
|
| 866.2 | | CASEE::CLEOVOULOU | Marios Cleovoulou | Sun Jun 04 1989 12:17 | 33 |
| > Not having seen this elsewhere, what can we do if the process does "go
> away" ?
First of all, submit a QAR/SPR. Processes/applications are definitely
not supposed to just "go away"! You will find "evidence" as to why it
has done so in the output stream of the process (e.g. under "Work in
Progress" if it was started with FileView, in the terminal window if it
was $ RUN, in the log file if it was $ SPAWN/NOWAIT/OUT=..., etc).
Submit this "evidence" with your QAR/SPR.
> ...what can the user (with only OPER privs) do to bring it back ?
The same as the user did to start it in the first place. (You have the
answer for DECterms and FileViews created on login). You won't need
any more privs to restart it that you did in the first place (none for
most apps anyway).
> The "reasons" for these answers would be good to know, is so, that if
> the VS2000 and DECwindows votes to halt one of the processes (voted in
> secret) due to some constraint; then the blissfull operator might stand
> a chance of trading a non-stopped lower level priority process ...
The system doesn't really work that way. If you are having problems
(re)starting applications then the system parameters and/or user's
quotas are probably badly set up. If you are starting things from
FileView it will, generally, tell you why it can't start something if
that is the case. The release notes cover various parameters that
affect the number etc of processes that you can get going.
Regards,
Marios
|
| 866.3 | QAR FT Product. ? Joke, right ? (Also "If" ..) | WILKIE::EARLY | Bob Early CSS/NSG Dtn 264-6252 | Mon Jun 05 1989 10:54 | 57 |
| re: .2
QAR/SPR ?
The "DECWindows" are a FIELD TEST (to the best of my knowledge). How
do you QAR or SPR a "Field Test" product ?
This system who got "dumped" on me, by managerial default. The systems
probably came from IEG, so at that price, who should complain when it
doesn't work right, eh ?
If someone were to submit an SPR, it'll probably sit on someones desk
(like so many others) for awhile before being forwarded to the proper
support group ( caveat emptor). And then maybe (like VMS and RSTS), it
might get fixed someday in the future.
No, the better idea will be to let the product get sold "as is", and
let the customers submit CLDs. Then they'll get fixed "fer sure" and
lots faster.
Please don't tell me this is rotten attitude. I already know that.
My STRESS level constantly bounces up and down with DECwindows because:
1) It is not a reliable system.
- Loses processes without notice
- Reinitializes "default files" on Startup (losing data)
- Unable to easily define backup paramaters (to save defaults)
2) It is tedious to load/unload (QUIT SESSION/ ReLogin)
- Usually twice due to the LOGIN WINDOW
3) No USER manual was shipped with it (must be DEC product).
4) It reconfirms my belief the DEC (as a Corporation will not learn)
5) Customer Satisfaction only applies to major accounts.
6) Some of the Developers, while trying very hard to provide some help
to internal folks, don't seem to be getting enough help to try out and
reengineer a better solution.
If this were the FIRST product I got to use when it was new, it'd be a
nuisance at its worst.
If it were the ONLY DEC product that was shipped prior to optimization,
it'd be an inconvenience.
If I could feel confident, that once the product is debugged and 99%
operational (as spec'd) that it wouldn't be replaced by yet another
"advanced product", (i'd be out of my mind.)
Once, I had the chance to ensure that the products that got shipped not
only had manuals, but the manuals head "real indexes" in them, and
every customer got one.
Bob
|
| 866.4 | | CASEE::LACROIX | Gone with the wind | Mon Jun 05 1989 11:36 | 52 |
| Hey Bob, how about checking things before flaming like crazy to who
ever comes passing by?
> The "DECWindows" are a FIELD TEST (to the best of my knowledge). How
> do you QAR or SPR a "Field Test" product ?
DECwindows V1 isn't in Field Test, thank god! It's shipping and has
been for MONTHS (part of VMS V5.1).
> If someone were to submit an SPR, it'll probably sit on someones desk
> (like so many others) for awhile before being forwarded to the proper
> support group ( caveat emptor). And then maybe (like VMS and RSTS), it
> might get fixed someday in the future.
You won't believe it, but the SPR process is being closely looked out,
from what I've heard. I don't know about all other engineering
organizations, but when we get an SPR here in Valbonne, we immediately
stop whatever we were doing and start working right on the spot on the
SPR's answer, and we then schedule time for a fix if one is needed. I
have reasons to believe that this is the case for most if not all
engineering organizations. If you don't subit a QAR, how are we
supposed to find ALL bugs??!
> 1) It is not a reliable system.
> - Loses processes without notice
> - Reinitializes "default files" on Startup (losing data)
> - Unable to easily define backup paramaters (to save defaults)
Try to pin point what the problem is, and submit a QAR.
> 3) No USER manual was shipped with it (must be DEC product).
Bull. Have you heard of the following manuals: Overview of VMS
DECwindows, VMS DECwindows User's Guide, VMS DECwindows Desktop
Applications Guide, etc...
> 6) Some of the Developers, while trying very hard to provide some help
> to internal folks, don't seem to be getting enough help to try out and
> reengineer a better solution.
Have you heard about compromising? About money being a scarce resource?
No one will ever claim that DECwindows V1 is perfect, nor that V8 will
be. In fact, I'll agree that DECwindows has hundreds or thousands of
bugs; so what? DECwindows is millions of lines of source code, and one
bug per thousand line would be a pretty good ratio ;-) Name one
software engineering project in Digital that was started in the past 10
years, and which had the scope, size and complexity of DECwindows. How
many did you find?
Denis.
|
| 866.5 | Oh, it just gets funnier the more we laugh!! | WOODRO::EARLY | Bob Early CSS/NSG Dtn 264-6252 | Thu Jun 15 1989 12:01 | 107 |
|
re: .4
> QAR/SPR
Ok, what is the process to get a QAR form (or SPR Form) ? WHats the
difference between them ?
> The "DECWindows" are a FIELD TEST (to the best of my knowledge). How
> do you QAR or SPR a "Field Test" product ?
>> DECwindows V1 isn't in Field Test, thank god! It's shipping and has
>> been for MONTHS (part of VMS V5.1).
You mean its been bundled as part of VMS 5.1 ?
> If someone were to submit an SPR, it'll probably sit on someones desk
> (like so many others) for awhile before being forwarded to the proper
> support group ( caveat emptor). And then maybe (like VMS and RSTS), it
> might get fixed someday in the future.
Hey, sorry if this sounds bad to you. Its based on "personal knowledge"
of SPRs that were submitted. Even some customers echo the sentiment
that SPRs are a "Black Hole" (sort of like ECOs to some hardware
plants.
>> You won't believe it, but the SPR process is being closely looked out,
>> from what I've heard.
Sure, I can believe that memos have been issued, plans have been
generated, and commitments have been signed up to. But, and I just
feel that you can catch this one, many SW groups can easily show
that the reported problem isn't caused by "their" process, but due
to the "mis-coding" of an add-on product (DECwrite, DECW$Calendar,
DECW$NOTEPAD, etc). And golly gee folks, we can't do THEIR job
now, can we ? Oh yes, I've heard this one before, too many times.
I don't guess anyone will find all the bugs anyway. Its just that
when the process crashes, and the "jobsave.LOG" file wasn't journaled;
there's nothing to examine. I'm a user, not a SW engineer.
> Please don't tell me this is rotten attitude. I already know that.
> 2) It is tedious to load/unload (QUIT SESSION/ ReLogin)
> - Usually twice due to the LOGIN WINDOW
> 3) No USER manual was shipped with it (must be DEC product).
>> Bull. Have you heard of the following manuals: Overview of VMS
>> DECwindows, VMS DECwindows User's Guide, VMS DECwindows Desktop
>> Applications Guide, etc...
No, no bull. The only manual included was the 'installation' guide,
an NETwork set up guide. The rest of the manual had to be separately
ordered, and the backlog was several weeks.
(EK-V2000-SG-001, EK-NETAC-UG-003
> DECwindows is millions of lines of source code, and one
> bug per thousand line would be a pretty good ratio ;-) Name one>
One bug per one thousand lines of code, where one line is one
instruction, is about par for some programmers, isn't it, at least
from <whose> perspective, for a RELEASED product.
1,000,000 / 1000 = 1000 bugs per million instructions. Sounds high.
Did the same group do the launch program for the Space Shuttle.
At 10MPS, thats about a 1/10 sec reliability, isn't it ?
The Challenger ?
The point isn't how clever we can with words. The point is that
VMS 5.1 with DECwindows is supposed to be used IN PLACE OF a very
reliable VMS that didn't give me 1/10th all this STRESS and PAIN
and anguish.
VMS 4.7 didn't pretend to support anything other than itself, so
noone used it for anything other than an elecronic interface to
host files.
What the previous methodology did was to create BACKUP files in
the event the CPU got lost.
Right now "my" VS2000 with Disk is sitting there fat, dumb, happy,
and non-functional because "something" happened to the boot_node,
and this piece of "hardware" hasn't enough intelligence to give
a "warning message" that "its broke, folks".
14 Mb memory, a 40MB disk, and not enough functionality to say "I'm
broke, please fix me".
The funny part is: We'll look back on this next year and laugh
about it, when we consider how trivial it is in relationship to
the NEXT level of <os>.
Naw, i'm not mad. just frustrated to tears that my dinky little
Rainbows' still chugging away its electrons, limited in functionality,
but then .. its not supposed to work miracles. its only a computer.
;^)
Bob
|
| 866.6 | | PSW::WINALSKI | Careful with that VAX, Eugene | Fri Jun 16 1989 21:22 | 14 |
| RE: .5
If you have problems that you want to see fixed, the way to report them to the
developers is via the QAR system. SET HOST to node TRIFID. Log in to username
QAR_INTERNAL, password QAR. Select the DECWINDOWS-IFT database. Use the ENTER
command in the QAR system to post a new problem report.
QARs don't necessarily get answered promptly, but they are seen and the problems
are tracked so that they don't drop on the floor. Grousing in notes conferences
accomplishes nothing--the developer responsible for fixing the problem may
never see your complaint.
--PSW
|
| 866.7 | Ok, Thanks for the pointer. | WOODRO::EARLY | Bob Early CSS/NSG Dtn 264-6252 | Tue Jun 20 1989 09:05 | 27 |
| Thanks for both:
The information on "How to" ...
and not beating up on me.
One more question:
Is this QAR just for DECwindows or all DECW$ processes (applications)?
Part B of Question #2
What file should I save after a "crash" ? What is the name of it?
Thers is one, called something SM$ (i don't believe it. i just lost
them "AGAIN" !).
Ok, IF and WHEN i get thes DECwindows BACK .. AGAIN .. What file
need I need to recover to ensure the <whoever> can track the WHY
the doggone DECwindows "just plain died".
Bob Early
(this is so da***m frustrating ....)
|
| 866.8 | Ok, here' the FAIL.lOG and the DECW$SM.LOG. Where? | AHIKER::EARLY | Bob Early CSS/NSG Dtn 264-6252 | Tue Jun 20 1989 12:53 | 101 |
|
Ok, I got this info from the decw$sm.log file, and another from
a FAIL.LOG.
What do we next to try to figure out what went wrong ?
I mean, this little darling crashes DECwindows on an average of
3 times a day (or more); and the system (VS2000) about 2- 3 times
per week.
*Assuming* i may be causing the condition to exist (as a user);
what can we do to prevent it ?
Bob Early
%DCL-S-SPAWNED, process EARLY_1 spawned
%DCL-S-SPAWNED, process EARLY_2 spawned
%DCL-S-SPAWNED, process EARLY_3 spawned
%DCL-W-ACTIMAGE, error activating image SYS$SYSTEM:DECW$WINDOW
-CLI-E-IMAGEFNF, image file not found STATE$DUA0:[SYS11.SYSCOMMON.][SYSEXE]DECW$WINDOW.EXE;
EARLY job terminated at 19-JUN-1989 15:20:35.16
Accounting information:
Buffered I/O count: 27 Peak working set size: 223
Direct I/O count: 9 Peak page file size: 1965
Page faults: 296 Mounted volumes: 0
Charged CPU time: 0 00:00:01.02 Elapsed time: 0 00:00:07.29
Session Message: Started DECterm controller
%SYSTEM-F-ACCVIO, access violation, reason mask=00, virtual address=00000000, PC=000104BC, PSL=0BC00
Improperly handled condition, image exit forced.
Signal arguments Stack contents
Number = 00000005 0001035A
Name = 0000000C 00000000
00000000 200C0000
00000000 7FF373F4
000104BC 7FF373DC
0BC00000 000FA551
001E406D
001E406D
00000000
27FC0000
Register dump
R0 = 07007FFF R1 = 00000000 R2 = 00000000 R3 = 000100C8
R4 = 00000000 R5 = 00000000 R6 = 00000000 R7 = 00000003
R8 = 00000F8B R9 = 7FF37A90 R10= 001A3B24 R11= 0000FFFE
AP = 7FF37374 FP = 7FF37334 SP = 7FF373B0 PC = 000104BC
PSL= 0BC00000
EARLY job terminated at 19-JUN-1989 15:38:05.25
Accounting information:
Buffered I/O count: 115 Peak working set size: 1000
Direct I/O count: 49 Peak page file size: 5562
Page faults: 6763 Mounted volumes: 0
Charged CPU time: 0 00:00:31.21 Elapsed time: 0 00:17:35.34
XIO: non-translatable vms error code: 0x2DBA002, vms message:
%decw-e-cnxabort, connection aborted
%XLIB-F-IOERROR, xlib io error
EARLY logged out at 19-JUN-1989 16:15:40.59
=============================
FAIL.LOG
�
%SYSTEM-F-PAGRDERR, page read error, reason mask=00, virtual address=0000CDDC, P
C=0000CDDD, PSL=03C00000
Improperly handled condition, image exit forced.
Signal arguments Stack contents
Number = 00000005 0000C051
Name = 00000444 000200FF
00000000 7FF37634
0000CDDC 7FF37632
0000CDDD 00000000
03C00000 010E000C
0000BFC8
010E000F
0000BFD4
00070000
Register dump
R0 = 00000007 R1 = 7FF37634 R2 = 7FF384C4 R3 = 00000001
R4 = 0000004D R5 = 00000000 R6 = 00000003 R7 = 000114F0
R8 = 00001993 R9 = 00011480 R10= 0000698C R11= 000012A4
AP = 7FF375D0 FP = 7FF37590 SP = 7FF3760C PC = 0000CDDD
PSL= 03C00000
|
| 866.9 | Looks like hardware! | CASEE::CLEOVOULOU | Marios Cleovoulou | Tue Jun 20 1989 13:15 | 11 |
| > %SYSTEM-F-PAGRDERR, page read error, ...
Well, this looks like your disk is having problems. A quick look in
the System Messages and Recovery Procedures manual indicates that
$ ANAL/DISK/READ_CHECK would be useful at this point. A call to
field-service may be more appropriate than a QAR here!
Regards,
Marios
|
| 866.10 | Was it softare or a flaky disk ?? | WOODRO::EARLY | Bob Early CSS/NSG Dtn 264-6252 | Thu Jun 22 1989 09:48 | 21 |
| re: .9
''''''''''''''''''.....................---------------------------------------
############%%%%%%%%%%%%%%%%%%%%$$$$$$!!!!!!
Thnks.
The new question is now: When the "Disk Error bit is set", is there a
recovery procedure to get the data from the disk, for its final backup?
(don't tell me about daily backups. noone told me about doing them
before it crashed, so i don't want to hear about them now).
Bob
-btw- case in point about dec equipement from IEG (another case
in point).866.3
|
| 866.11 | Crows are tough eating, but the problem may be solved. | WILKIE::EARLY | Bob Early CSS/NSG Dtn 264-6252 | Wed Jul 05 1989 10:02 | 27 |
| Re: .1-.9
What was that quote now "Keep your words short and sweet so if you
need those words to eat".
1) IFS replaced the harddisk and the expansion module.
2) Local LAT mis had began backups on some systems alphabetically,
so of the few systems backed up, mine was one .
3) The systems does have a WARRANTY contract on it, so after i called
the CSC, and after IFS understood my needs; the system got fixed.
4) The new dusk is about 1/2 as noisy as the old disk.
5) Noone else expeerienced the same level of problems (with DECW$
as i have during this phase).
6) Stress does one heck of a number on you when you feel "rushed".
7) There is no excuse for hurting good friends, peers, or people
trying to help you.
Thanks,
Bob
|