Re: Automated report: NetBSD-current/i386 test failure

[Joerg Sonnenberger <joerg%bec.de@localhost>]

On Sat, Apr 18, 2020 at 10:32:18AM +0300, Andreas Gustafsson wrote:
> The NetBSD Test Fixture sent three reports listing the following
> groups of commits, respectively:
> 
> >    2020.04.16.14.39.58 joerg src/lib/libc/gen/pthread_atfork.c,v 1.13
> >    2020.04.16.14.39.58 joerg src/libexec/ld.elf_so/rtld.c,v 1.204
> >    2020.04.16.14.39.58 joerg src/libexec/ld.elf_so/rtld.h,v 1.139
> >    2020.04.16.14.39.58 joerg src/libexec/ld.elf_so/symbols.map,v 1.3
> 
> >    2020.04.16.18.20.46 msaitoh src/sys/dev/pci/pcidevs,v 1.1406
> >    2020.04.16.18.21.12 msaitoh src/sys/dev/pci/pcidevs.h,v 1.1394
> >    2020.04.16.18.21.12 msaitoh src/sys/dev/pci/pcidevs_data.h,v 1.1393
> >    2020.04.16.18.32.29 msaitoh src/sys/dev/pci/ichsmb.c,v 1.67
> >    2020.04.16.18.51.47 pgoyette src/share/man/man4/man4.x86/imcsmb.4,v 1.8
> >    2020.04.16.18.56.04 pgoyette src/share/man/man4/man4.x86/imcsmb.4,v 1.9
> >    2020.04.16.19.23.50 bouyer src/sys/arch/xen/xen/Attic/xen_clock.c,v 1.1
> 
> >    2020.04.16.15.47.19 christos src/external/gpl3/binutils/dist/ld/emultempl/elf.em,v 1.2
> >    2020.04.16.15.58.13 jdolecek src/sys/external/mit/xen-include-public/dist/xen/include/public/io/blkif.h,v 1.2
> >    2020.04.16.16.38.43 jdolecek src/sys/arch/xen/xen/xbd_xenbus.c,v 1.116
> >    2020.04.16.17.18.27 nat src/sys/dev/ic/rtwnreg.h,v 1.3
> >    2020.04.16.17.18.27 nat src/sys/dev/usb/if_urtwn.c,v 1.86
> 
> The latter two reports are spurious, and the commits listed in them
> have nothing to do with the breakage.  The reason for the spurious
> reports is that a large number of t_ptrace_wait* test cases started
> failing with the commit listen in the first report, but are not failing
> in every run.  Tests that happened to pass in the first run and fail
> four times in a row after that got reported one commit too late, etc.

AFAICT the tests are bad. They fail significantly more often when
compiled with optimisations than without, strongly suggesting race
conditions involved. The original change here certainly changes the
timing and I am aware of one potential bug in it, but that bug is most
definitely not exercised by the test cases.

Joerg