On Wed, Dec 17, 2014 at 05:19:21PM +0000, Chris Wilson wrote: > char *new, *old; sigset_t block_signals, save_signals; > asprintf(&new, > "(%s:%d) %s%s%s: %s", > program_name, getpid(), > (domain) ? domain : "", > (domain) ? "-" : "", > igt_log_level_str[level], > line); sigfillset(&block_signals); pthread_sigmask(SIG_SETMASK, &block_signals, &save_signals); > > pthread_mutex_lock(&log_buffer_mutex); > old = log_buffer.entries[log_buffer.end]; > log_bufffer.entries[log_buffer.end] = new; > if (++log_buffer.end == log_buffer.start) > log_buffer.start++; > pthread_mutex_unlock(&log_buffer_mutex); pthread_sigmask(SIG_SETMASK, &save_signals, NULL); > free(old); That will at least stop SIGARLM and friends firing whilst we poke around inside log_buffer. It might make a test result unstable, but will prevent that same instability from causing a lockup of igt. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
