xet_robust_list--pidns-translation.gen.test sometimes failing

[Sven Schnelle]

Hi,

we're seeing the following test failure in CI with strace:

not ok 1035 - xet_robust_list--pidns-translation.gen.test
  ---
    testlog: |
      raw log output
      ============================================
      
      --- exp	2022-03-01 04:04:10.107036332 +0100
      +++ out	2022-03-01 04:04:10.167036332 +0100
      @@ -4,7 +4,8 @@
       2463824 get_robust_list(2463824, [0x3ff82ce8fe8], [24]) = 0
       2463824 +++ exited with 0 +++
       2463826 get_robust_list(0, NULL, NULL) = -1 EFAULT (Bad address)
      -2463826 get_robust_list(2 /* 2463826 in strace's PID NS */, [0x3ff82cf8c00], [24]) = 0
      +2463826 get_robust_list(2 /* 2463826 in strace's PID NS */,  <unfinished ...>
      +2463826 <... get_robust_list resumed>[0x3ff82cf8c00], [24]) = 0
       2463826 set_robust_list(0x3ff82ce8fe8, 24) = 0
       2463826 get_robust_list(2 /* 2463826 in strace's PID NS */, [0x3ff82ce8fe8], [24]) = 0
       2463826 +++ exited with 0 +++
      xet_robust_list--pidns-translation.gen.test: failed test: ../../src/strace --decode-pids=pidns -f -a24 -e trace=get_robust_list,set_robust_list ../xet_robust_list--pidns-translation output mismatch
      FAIL xet_robust_list--pidns-translation.gen.test (exit status: 1)

This happens only sometimes, and it looks to me like the cause is just
that the system call is displayed with the unfinished/resumed info.

I haven't found an option in strace to disable the splitting of the
syscall when another thread is interferring, and i'm not sure whether
the strace test frame work should handle this. Any thoughts?

Thanks
Sven