apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Massimo Manghi <mxman...@apache.org>
Subject Re: global mutex lock waits forever (SOLVED)
Date Wed, 09 Mar 2011 12:00:03 GMT
 In a private message Christopher Key gave the correct hint
 to solve the problem: the program was missing the appropriate
 initialization every APR based application must do by calling
 apr_app_initialize (and apr_terminate upon application exit).

 As an apology for my hapless inability to understand I
 would like to point out that at least another APR module
 I personally tested (shared memory), works pretty
 well on Linux without any apr_initialize() call at all.

  -- Massimo

 On Wed, 09 Mar 2011 10:54:48 +0100, Massimo Manghi wrote:
> I'm perhaps not very polite asking on this list without
> waiting the usual lurking time, but I've been googling
> this problem for a while yesterday and I did not find out
> anything about it. Still the problem is very clear:
> I wrote a sort of 'hello world' program where a global
> mutex is created to synchronize access to a resource
> shared between a parent and a child process created with a
> Unix 'fork' call.
> The child *should* gain the lock first, as the parent is given
> a 2 seconds wait before attempting to lock the mutex. Then
> it should release the lock giving way to the parent. That's all.
> Simply the calls to apr_global_mutex_[lock|trylock|unlock]
> wait forever and never return.
> I also thought the apr_global_mutex_create returned an implicitly
> locked mutex and tried to unlock it. Also this call would block
> forever. Where did I get it wrong?
> I asked also other people who are using global mutexes in
> their projects and they observed the same problem when they
> run my program and could not give an explanation.
> The program is so simple that I dare attach it here. It is
> a way to become familiar with tools that I hope will be used to
> design and implement a new software component to be developed
> for Apache Rivet, if this can work as an excuse for my
> intrusiveness.
> thanks for any help.
>  -- Massimo Manghi

View raw message