On 3/5/2014 2:51 PM, Gedare Bloom wrote: > Joel, > I suspect you need more changes in confdefs.h to put the keys in the > configuration and debug configurations structures, and to accurately > estimate the workspace memory usage, in case POSIX_KEYS are used > without having posix enabled. Ahh... yep... forgot both those. Not surprising since this turned out to be a lot of simple changes spread out. > Nothing else seemed obviously wrong at a quick read through. > Considering you updated the .scn files, I take it you have run the > psxkey tests with this change. Yep. Both with POSIX enabled and disabled.
I will fix that in condefs.h and shoot it again. > Gedare > > On Wed, Mar 5, 2014 at 2:36 PM, Joel Sherrill <joel.sherr...@oarcorp.com> > wrote: >> Hi >> >> As discussed in various threads and private conversations, task variables >> are broken by design on SMP systems. POSIX keys are a standard way >> to have SMP safe per task data. But currently they are only available >> when --enable-posix is specified. >> >> This rather tedious patch enables them all the time. It touches the >> configuration, code, and reworks the tests to use Classic tasks. >> >> The next step is to eliminate the use of task variables in the >> cpukit code. That code is unsafe for SMP systems. >> >> OK to apply? >> >> -- >> Joel Sherrill, Ph.D. Director of Research & Development >> joel.sherr...@oarcorp.com On-Line Applications Research >> Ask me about RTEMS: a free RTOS Huntsville AL 35805 >> Support Available (256) 722-9985 >> >> >> _______________________________________________ >> rtems-devel mailing list >> rtems-devel@rtems.org >> http://www.rtems.org/mailman/listinfo/rtems-devel >> -- Joel Sherrill, Ph.D. Director of Research & Development joel.sherr...@oarcorp.com On-Line Applications Research Ask me about RTEMS: a free RTOS Huntsville AL 35805 Support Available (256) 722-9985 _______________________________________________ rtems-devel mailing list rtems-devel@rtems.org http://www.rtems.org/mailman/listinfo/rtems-devel
