>How I got around this was to apply Thomas Schwinge's fix_have_kernel_resources >branch to libpthread. Basically, the first thread structure allocated is not >zeroed out, >so have_kernel_resources is some random non-zero number, and thus the thread >neither gets a wakeup port, nor does the kernel_thread member get initialized. >There may be other issues to work on.
My bad. There are. This might be a problem in the Debian patches, as this specific problem disappeared for me when fix_have_kernel_resources was the ONLY patch I was applying against libpthread. I forgot about this issue because it disappeared for me. That was three months ago. On the subject of fix_have_kernel_resources: what further work can I do to get this integrated? I can split the predicates. I can also go into a lengthy rant as to why, to the best of my understanding, both resources are treated in an atomic manner, with the exception of one instance (which I have commented on before). I am not going to fix something that I don't see as broken without guidance. Thomas D
