Skip Menu |

This queue is for tickets about the threads CPAN distribution.

Report information
The Basics
Id: 18419
Status: resolved
Priority: 0/
Queue: threads

Owner: Nobody in particular
Requestors: jdhedden [...]

Bug Information
Severity: Important
Broken in: 1.19
Fixed in: (no value)

Subject: pthread_attr_setstacksize failures under Linux
There have been a number of failed test reports with the following error message: Thread creation failed: pthread_attr_setstacksize(2097152) returned 22 (Note: The error message is descriptive - the actual function call is below). This shows that the pthread_attr_setstacksize function was called with the stack size of 2MB (2*1024*1024), and that the function returned EINVAL. The architectures (and failure reports) under which this has been reported so far include: linux lappy 2.4.22 (Perl 5.8.7 with use64bitint=define) linux penguin 2.4.26 (Perl 5.8.6 and 5.9.3 with use64bitint=define) linux promomail 2.4.27-2-386 (Perl 5.8.8 with use64bitint=undef) debian 3.1 stable Checking the source code for the pthread_attr_setstacksize function in glibc shows that the only cause for the EINVAL error code is too small of a stack size: /* Catch invalid sizes. */ if (stacksize < PTHREAD_STACK_MIN) return EINVAL; Usually that minimum is 16K, however, the threads module code checks against whatever PTHREAD_STACK_MIN is set to, so 2MB should be valid. Within the threads module code, stack size is carried as an IV (64-bit int): typedef struct ithread_s { ... IV stack_size; } ithread; IV stack_size; And the call to pthread_attr_setstacksize casts it to size_t: # ifdef _POSIX_THREAD_ATTR_STACKSIZE /* Set thread's stack size */ if (thread->stack_size > 0) { rc_stack_size = pthread_attr_setstacksize(&attr, (size_t)thread->stack_size); } # endif The fact that one architecture which doesn't use 64-bit ints (linux promomail 2.4.27-2-386 (Perl 5.8.8 with use64bitint=undef)) and yet still reports the failure shows that the problem is not related to casting the IV to size_t. This bug is 'stalled' until someone can shed some more light on the issue.

This service is sponsored and maintained by Best Practical Solutions and runs on infrastructure.

Please report any issues with to