Kelly Robbins wrote:
> Roger,
>
> How much work is considered a lot? There are 4 lines to the signal
> handler... 1 to print the fact that I made it there :), 1 to put together
> an output string, and 2 calls to a procedure (about 10 lines) to spit out
> some html output.
>
> Kelly
According to what i have read, that may be too musch. Mucking around with memory
structures in a sig
handler can confuse perl if the changes made inside the sig handler confict with
what perl was
doing befor it caught the sig (updating or deleting that memory structure - I am not
a guts person
so I proboly said that wrong...) anyway, the book recomends just throwing a die() or
seting a global
flag in the sig handler and then checking for that flag in the "if ($@)" block and
doing all your
real work there. You might try moveing the warning, the procedure calls, and the
string creation
into the "if ($@)" block and seeing if that make stuff more stable. You might also
try your code on
different versions of perl - I think 5.6.1 addressed some sig handler issues, but I am
not sure.
Roger
>
>
> On Tue, 5 Mar 2002, Roger Grayson wrote:
>
>
>>Kelly Robbins wrote:
>>
>>>Tim,
>>>
>>>This is exactly the example that I used. I can get the program to run for
>>>a few iterations, and handle a few timeouts, but it eventually seg
>>>faults. :(
>>>
>>>I'm including the output generated from print statements in my script. It
>>>shows a few iterations, a few timeouts, then a seg fault. I can get the
>>>signal to work, but the seg fault over time is what I'm concerned about.
>>>Any suggestions for this?
>>>
>>>At CheckTNSPing
>>>At CheckDiskSpace
>>>Timeout while performing DiskSpace alert
>>>Sleeping for 5 seconds starting @ 2002/03/05 17:01:47.
>>>At CheckTNSPing
>>>At CheckDiskSpace
>>>At CheckSessionInfo
>>>At CheckInvalidRows
>>>At CheckPermissions
>>>Timeout while performing Permissions alert
>>>Sleeping for 5 seconds starting @ 2002/03/05 17:02:08.
>>>At CheckTNSPing
>>>At CheckDiskSpace
>>>Timeout while performing DiskSpace alert
>>>Sleeping for 5 seconds starting @ 2002/03/05 17:02:22.
>>>At CheckTNSPing
>>>At CheckDiskSpace
>>>At CheckSessionInfo
>>>At CheckInvalidRows
>>>At CheckPermissions
>>>At CheckMaxParameters
>>>At CheckObjectExtents
>>>Timeout while performing ObjectExtents alert
>>>Sleeping for 5 seconds starting @ 2002/03/05 17:02:43.
>>>At CheckTNSPing
>>>At CheckDiskSpace
>>>zsh: 11968 segmentation fault /usr/local/bin/perl5
>>>/usr/vendor/oracle/UU-DBA/cr
>>>on/../bin/OracleMonitor -s
>>>
>>>
>>>On Tue, 5 Mar 2002 [EMAIL PROTECTED] wrote:
>>>
>>>
>>>
>>>>Kelly,
>>>>
>>>>I think that what you want to achieve can be done with the alarm() function.
>>>>If you use it in an eval block, it won't kill the program. Here's a simple
>>>>example (from "Perl: The Complete Reference" by Martin Brown, pg 455):
>>>>
>>>> print "Your name is? :\n";
>>>> eval
>>>> {
>>>> local $SIG{ALRM} = sub { die "Timeout" };
>>>> alarm(10);
>>>> $name = <STDIN>;
>>>> chomp $name;
>>>> alarm 0;
>>>> };
>>>>
>>>> if ($@ and $@ =~ /Timeout/) { $name = 'Anonymous' }
>>>> print "Hello $name!\n";
>>>>
>>>>alarm(10) sets the alarm for 10 seconds. alarm(0) turns it off. The special
>>>>variable $@ aka $EVAL_ERROR will be empty if the eval execute without error,
>>>>otherwise it holds the "die" message.
>>>>
>>>>There is also a good discussion of this stuff in the O'Reilly book
>>>>"Programming Perl".
>>>>
>>>>Hope this helps,
>>>>
>>>>Tim Helck
>>>>
>>>>-----Original Message-----
>>>>From: Kelly Robbins [mailto:[EMAIL PROTECTED]]
>>>>Sent: Tuesday, March 05, 2002 12:17 PM
>>>>To: [EMAIL PROTECTED]
>>>>Subject: Signal to timeout a long running query
>>>>
>>>>
>>>>
>>>>I'm trying to incorporate a signal in a database monitoring tool that we
>>>>have written which runs "forever". I would like to be able to break out
>>>>of a long running query, but I would also like the script to continue
>>>>running.
>>>>
>>>>I have been successful writing a signal to quit a query that is running to
>>>>long, and the script will then continue, but the script eventually seg
>>>>faults over time. Is there any way to prevent the seg faults? If not, is
>>>>there a better way of imposing a timeout on my queries?
>>>>
>>>>Kelly
>>>>
>>>>
>>>>
>>>>
>>Kelly,
>>
>>How much work is your signal handler doing? Lincoln mentions in "Network Programming
>with Perl" that
>>sig handlers can cause segfaults if you do a lot of work in them. So if you are
>doing more than just
>>calling die() or setting a global var in your sig handler that could be the problem.
>It might also
>>matter which version of perl you are using.
>>
>>Roger
>>
>>--
>>There are lies, there are damned lies, and there are the power point presentations
>you make up to
>>explain your code.
>>
>>
>
>
--
There are lies, there are damned lies, and there are the power point presentations you
make up to
explain your code.
