FYI: Your code appears to function as designed, without errors, when run under 5.8.5 and win32.
I think dave_the_m called it right when he suggested you upgrade. Threads were still quite flakey in 5.8.1.
Is there a particular reason for using a single scalar to guard both your shared hashes rather than locking the hash itself (or the element being accessed) individually?
In reply to Re^2: Problems with a thread-pooled tcp server
by BrowserUk
in thread Problems with a thread-pooled tcp server
by ph713
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |