On 28.03.19 17:13, Stephen D. Cohen via Xenomai wrote:
Folks,I noted a couple of issues with rt_task_inquire and wanted to share them. I will put together a patch later <makes hand rolling motion>, but wanted to let folks know sooner rather than later. The documentation suggests that the info structure can be NULL. This is simply not true, as the actual code dereferences the pointer without checking. Thus a null info pointer will cause an exception and the thread will die an untimely (though outwardly quiet) death. I also note that the rt_task_inquire routine has an exit path that keeps the thread object in task control block locked. This is also a problem. There should really be a put_alchemy_task before the "goto out;" after the threadobj_stat call. I noted that this "keeps the tcb->thobj locked" issue occurs in a few of the other rt_task routines as well. A careful review is in order and I will do it when I submit the patches for rt_task_inquire.
Thanks for the report! Sounds reasonable - looking forward to patches. Jan -- Siemens AG, Corporate Technology, CT RDA IOT SES-DE Corporate Competence Center Embedded Linux
