Mailing-List: contact dev-help@apr.apache.org; run by ezmlm
Precedence: bulk
To: dev@apr.apache.org
Path: not-for-mail
From: Ian Holsman <ianh@apache.org>
Newsgroups: server.http.apr.dev
Subject: Re: pool debug question.
Date: Thu, 05 Sep 2002 08:16:09 -0700
Organization: Holsman.NET
Lines: 45
Message-ID: <3D777539.3070204@apache.org>
References: <3D758105.5040906@apache.org>
 <JLEGKKNELMHCJPNMOKHOEEBBGMAA.striker@apache.org>
NNTP-Posting-Host: int.holsman.net
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit
NNTP-Posting-Date: Thu, 5 Sep 2002 15:16:08 +0000 (UTC)
To: Sander Striker <striker@apache.org>
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US;
 rv:1.1) Gecko/20020826

Sander Striker wrote:
>>From: Ian Holsman [mailto:ianh@apache.org]
>>Sent: 04 September 2002 05:42
> 
> 
>>hi all.
>>
>>I've got a piece of multi-threaded code which runs fine
>>until it gets to the apr_terminate, where it segfaults in
>>apr_allocator_destroy.
>>
>>any hints on what is going on ?
> 
> 
> apr_initialize calls apr_pool_initialize which creates a
> global allocator which is used for all pools (unless overridden).
> The global allocator is protected by a mutex, which is allocated
> out of the global pool (also created by apr_pool_initialize).
> 
> apr_terminate calls apr_pool_terminate which in turn destroys
> the global pool.  The global pool is the owner of the global
> allocator and thus destroys the global allocator after its own
> destruction.  The global pool cleanup _should_ already have
> taken care of the destruction of the mutex and the resetting of
> the mutex field of the global allocator.
> 
> 
>>the test program is creating and destroying thousands of pools
>>but is not messing at all with allocators
> 
> 
> Are you sure you are not calling apr_terminate before all threads
> have finished?  If you are we have a nice thing to debug :(
> 
This could be the problem, as the program is creating/exiting threads
several times. but I am 99.9% sure that they are all dead before I 
terminate (this is what msvc's debugger shows me)

The code is on http://webperf.org/a2/prob/
it's the start of a dynamically sizing mpm (it spawns more threads
when the queue is getting bigger, longer term it will base it's decision 
on response times)
> Sander
>