T.R | Title | User | Personal Name | Date | Lines |
---|
46.1 | | GLIVET::DIAMOND | | Wed Aug 15 1984 18:44 | 6 |
| No way. You shouldn't be able to run an installed image with the debugger!
When you patch, you create a new image file and the old one remains installed.
To have privs in the new patched one, you'd have to re-install it. I don't
see how you even got the one with the debugger to work...
dave
|
46.2 | | ORPHAN::BRETT | | Wed Aug 15 1984 21:52 | 12 |
|
Privileged images SHOULD be linked /NODEBUG/NOTRACE. Failure to do this leads
to the aforementioned hole.
Unfortunately an earlier release of VMS had DISPLAY (precursor to MONITOR) with
this bug.
The real big thing to watch with privileged images is the affect logical names
can have on your program.
/Bevin
|
46.3 | | VAXUUM::DYER | | Thu Aug 16 1984 19:03 | 3 |
| I believe VMS V4 will not let you install an image that was linked
/DEBUG or /TRACE.
<_Jym_>
|
46.4 | | GLIVET::DIAMOND | | Fri Aug 17 1984 09:06 | 4 |
| Well, by golly you're right. I just tried a debugged program with install
and it worked...
dave
|
46.5 | | ROYCE::KENNEDY | | Wed Aug 22 1984 11:40 | 19 |
| YOU DON'T NEED TO LINK TO THE DEBUGGER TO USE IT
================================================
Some time ago, I was called in to look at a third party package (INFO)
from a firm called HENCO. The package worked OK under the normal
terminal driver but failed under the NV: driver (X25 remote terminal).
I hacked at the package using RUN/DEBUG and lo and behold - I was in the
debugger! Soon enough, I was able to locate an illegal terminal function
code which was not being picked up by TT: but was by NV:. There was no
trace-back info, just a raw image, a dissassembling debugger and the
promise of a Pint if I fixed it.
Another interesting little point is that if you call LIB$SIGNAL with a
status code of SS$_DEBUG, you end up in the debugger ( I havn't checked
this yet). By using this together with assignments of DBG$INPUT and
DBG$OUTPUT, this could be a neat way of debugging detached processes.
Hugh.
|
46.6 | | ROYCE::KENNEDY | | Wed Aug 22 1984 15:39 | 10 |
| Quick postscript to the previous reply - it is /NOTRACEBACK that
prevents the debugger from comming in when you say RUN/DEBUG.
Unfortunately /TRACEBACK is invaluable for getting information
out of crashed programs, especially in Field Test software.
If the comment about V4 INSTALL is correct then this potential
hole will go. Pity - It could have been quite useful!
hugh.
|
46.7 | | GLIVET::DIAMOND | | Wed Aug 22 1984 11:32 | 4 |
| Useful for whom? The only people who need to hack around like that are people
who probably shouldn't be!
Dave
|
46.8 | Debugger fun in VMS V2 | MDVAX3::COAR | And your little dog, 2! | Fri Oct 09 1987 22:28 | 11 |
| I remember using the debugger on a normal, unprivileged, program
from a normal, unprivileged, account. The system manager was walking
by, and I showed her what happened if I did a DEPOSIT PSL=0 - the
formatted values displayed by EXAMINE PSL showed me to be in kernel
mode, and to have always been that way (as far as VMS was concerned).
She nearly had a coronary on the spot; I had to take pity on her
and show her that it was the STORED PSL I was munging, not the active
one, and that VMS would crap on the debugger if it tried to restore
the altered PSL value.
#ken :-)}
|