| Title: | DIGITAL UNIX (FORMERLY KNOWN AS DEC OSF/1) |
| Notice: | Welcome to the Digital UNIX Conference |
| Moderator: | SMURF::DENHAM |
| Created: | Thu Mar 16 1995 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 10068 |
| Total number of notes: | 35879 |
Is there some way within Digital UNIX to have exceptions (for example, C++ "throw"n exceptions) to cause some user-defined code to be executed as the stack is unwound? I want to achieve an equivalent to a traceback, under program control. For example, on VMS (and with Ada exceptions) it was possible to use the "secondary exception vector" defined in the context of the "Condition Handling Facility" for this. The idea is that a random piece of code can throw an exception with a string, e.g. throw "some explanatory text"; // at __LINE__ 6 in __FILE__ file9.cxx and I'd like to be able to write a handler an arbitary number of callers up the stack which could generate the following when caught: file6.cxx[6]: not handled file7.cxx[13]: not handled file8.cxx[23]: not handled file9.cxx[6]: some explanatory text Ideally, the catcher would just be another piece of C++ that uses the "catch" statement. It would be nice if this could potentially co-exist with threads. I only need it in V4.0B or later, earlier would be a bonus. Any help greatly appreciated! Thanks, Shaheed P.S. This is not X-posted in the C++ conference as I figured that this was the right place to start - if not, I'll take it there.
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 9250.1 | COL01::LINNARTZ | Mon Mar 24 1997 08:44 | 24 | ||
Hi Shaheed,
rgd exception, C++ and threads. There had been a cleanup between
V3.2 and V4.0 in that the thread package uses the libexc.so now
for threads exception, so this works fine. unfortunatly the C++
exception package shouldn't be mixed with the thread exception package,
(reason's are named in the c_plus_plus conference) under V4.0. This
cleanup is planned for the next major release.
regarding the stackdump, I'll sent you a mail later on. I've just asked
somebody to mail me the module which contains some functionality you've
requested. essentially, we're using the sigwaiter thread. If we send
the signal from outside, it takes our control structures sets a
variable, walks about all threads and sent a pthread_kill(SIGBUS)
to those threads. in case of the variable being set, we then dump
out the stacktrace (using virtual_unwind and PC adjustment) and unset
the variable and return.
It works in general. We've discovered two reasons until now where it
fails (one I have to report) the other also fails with othre OS's
(ex HP's stackdump). It also fails if you're using assembler routines
which aren't conformant to the calling standard, but this is a nice
sideeffect, as you'll see those and can clean em up.
Pit
| |||||
| 9250.2 | SMURF::DENHAM | Digital UNIX Kernel | Mon Mar 24 1997 09:03 | 4 | |
Not sure how applicable this is, but there's code going into
PTMIN (the V4.0D release) that provides stack trace routines.
Not sure of the details. I can get you a development contact
if you're interested...
| |||||