perl-modperl mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Michael McLagan" <>
Subject Re: MP 1.27, A 1.3.27, P 5.8.0 -- SEGV in XS_Apache_write_client
Date Thu, 01 May 2003 13:23:28 GMT
On Thu, 01 May 2003 12:56:53 +1000, Stas Bekman wrote:

>> We're still getting SEGVs.  They've migrated somewhat with the patches 
>> I've applied but they're now coming in:
>>       ap_soft_timeout() on line 1762 which is:
>>       ap_set_callback_and_alarm(timeout, r->server->timeout);   
>> When I examine *r I get back a structure with a whole pile of stuff, 
>> all of which is 0.  It's like the call in XS_Apache_print line 1795 
>> which is:
>>       r = sv2request_rec(ST(0), "Apache", cv);
>> is returing a block of memory with nothing (all zeros) in it.  I 
>> cant look at r or ST but I can see that cv is a structure with 3
>> fields which don't seem to have a particular type.
>what do you get when you print *cv? why can't you see r?

It says "No symbol r in current context" or something like that, I don't
have the specifics in front of me.  I've no idea why it doesn't see
some of the symbols but does see others and why it's only within mod_perl
code that I can't seem to look at symbols.  It knows line numbers and 
the like but not the local variables.

I don't remember exactly what it showed for *cv other than it was either
one of the structs that starts with sv_any or it was a long list of vars
that I couldn't make any sense of.

>Could it be some hardware problems? I do see occasional problems like this on 
>my machine which seems to have a faulty AMD CPU :( I can't even reliably run 
>Do you have this problem on a single machine or can you reproduce it elsewhere 

If I move the site, the problem moves with it.

>Since you are using RH and mp-1.27 which is used by so many people the 
>problems that you report are very unlikely to be a problem with mod_perl.

I disagree.  Even if it's something being done in the perl code we have
running on the server, there should be some sort of protection built 
into mod_perl that keeps the daemon from crashing.   Given that applying
patches to mod_perl has resolved some of the crashing (probably only 
relocating the problem each time I patch an occurance), that says it's
in that part of the system.  On top of that, it seems to occur in one
particular script at this point, although not at all consistantly.

>> I *NEED* to get this resolved.  I'm open to suggestions and/or someone
>> with more specific knowledge taking a poke at some online (IRC)
>> exploration of a crashed httpd.
>/join #modperl

I'm there now (9:22 AM EDT) and seem to be the life of the party.
Hopefully I'll see some signs of life and someone there will have 
some insight into this.


View raw message