Re: preparing for the next release

[Mike F. <va...@ge...>]

On Mon 11 Aug 2014 22:00:43 Dmitry V. Levin wrote:
> On Mon, Aug 11, 2014 at 01:38:42AM -0400, Mike Frysinger wrote:
> > sh fails detach-stopped in the way you describe:
> > $ cat tests/detach-stopped.log
> > Process 10130 attached
> > --- SIGSTOP {si_signo=SIGSTOP, si_code=SI_USER, si_pid=0, si_uid=0} ---
> > --- stopped by SIGSTOP ---
> > brk(0)                                  = 0x417000
> > brk(0x438000)                           = 0x438000
> > nanosleep({120, 0}, Process 10130 detached
> > 
> >  <detached ...>
> > 
> > State:  S (sleeping)
> > detach-stopped.test: failed test: tracee is not group-stopped after detach
> > 
> > should we just mark this as XFAIL on old kernels ?
> 
> If I only could find out what's the first kernel version that has no such
> issues, I'd certainly do it.

let's see what x86+qemu has to say for this test
2.6.34: skip
2.6.39.4: skip
3.0.101: skip
3.1.1: skip
3.2.62: skip
3.3.8: skip
3.4.91: pass
3.6.5: pass
3.10.49: pass
so i guess your update to check for PTRACE_SEIZE is working :)

i noticed that uio compiled fine (because i had a glibc that provided 
preadv/pwritev), but the unittest can abort at runtime (not entirely sure 
under what circumstances like kernel config):
$ cat tests/uio.log 
uio: uio.c:24: main: Assertion `pwrite(fd, buf, sizeof buf, offset) == 4' 
failed.
./uio.test: line 16:  1378 Aborted                 ./uio
uio.test: failed test: uio failed

strace of uio itself shows:
open("/dev/null", O_WRONLY|O_LARGEFILE) = 3
pwrite64(3, "\0\0\0\0", 4, 1004211379570065135) = -1 EFBIG (File too large)
-mike