On 09/29, Oleg Nesterov wrote: > > On 09/25, Oleg Nesterov wrote: > > > > IOW, every time we need to check whether we should generate/report > > the artificial SYSCALL_EXIT stop event, we must also check whether > > we need send_sigtrap(). > > > > With this patch we pass the new step-from-clone.c test. > > > > Further changes: > > > > - I guess, we should also send SIGTRAP when SINGLESTEP resumes > > the tracee from SYSCALL_EXIT stop. This is trivial, but I need > > to write the test-case first. > > Yes, with this change the kernel passes all tests plus the test-case > below. I wrote it because I lost the hope to understand the required > behaviour ;) > > I'll send the patch tomorrow (hopefully I'll also send ->siginfo changes).
No. I am not going to show this patch. Yes it is trivial and even works. But the code becomes too ugly. OK. I never argued, ->ev_array should die. But I have to admit: it is even worse that I thought. I'll try to remove it tomorrow and send the patches. Oleg.
