> Mostly okay, but there's too many duplicated test names.
> 
>       % cat gdb.sum | sort | uniq -c | sort -rn
>       8 PASS: gdb.base/sigbpt.exp: pass SIGSEGV for no breakpoints
>       7 PASS: gdb.base/sigbpt.exp: rerun to main
>       4 PASS: gdb.base/sigbpt.exp: continue to keeper for no breakpoints
>       3 PASS: gdb.base/sigbpt.exp: stepi fault for no breakpoints
>       3 PASS: gdb.base/sigbpt.exp: continue to breakpoint at fault for no breakpoints
>       2 PASS: gdb.base/sigbpt.exp: pass SIGSEGV for bp before segv
>       2 PASS: gdb.base/sigbpt.exp: pass SIGSEGV for bp before and at segv
>       2 PASS: gdb.base/sigbpt.exp: pass SIGSEGV for bp at segv
>   
> Can you make more of the test names more unique?

Oops, I thought I'd covered that.  Try the attached.

> Last, a typo:
> 
>   set test "Verify that SIGSEGV occures at the last STEPI insn"

Fixed.

> Also I'm getting four FAIL results (native i686-pc-linux-gnu,
> red hat 8.0, gcc 3.3.3, binutils 2.15, glibc 2.2.93-5-rh,
> kernel 2.4.8-14-rh).
> 
>   FAIL: gdb.base/sigbpt.exp: stepi out of handler for no breakpoints (executed fault insn)
>   FAIL: gdb.base/sigbpt.exp: stepi out of handler for bp before segv (executed fault insn)
>   FAIL: gdb.base/sigbpt.exp: stepi out of handler for bp at segv
>   FAIL: gdb.base/sigbpt.exp: stepi out of handler for bp before and at segv
> 
> Is that a fault of the system under test, or a fault of the test script?
> Your message implies the former (because it passes with a hacked kernel),
> but I just want to check.

"yes", so far I've identified the following:

- h/w single-step #1 - kernel
The kernel is issuing two-for-one instructions when single stepping a 
system call.

- h/w single-step #2 - kernel
The kernel is continuing, instead of single-stepping, the sigreturn 
system call.

- s/w single-step #1 - gdb
GDB's OSABI software single-step code doesn't know how to single-step 
sigreturn.

We, GDB, can work around the kernel bugs by using software single step. 
  But that means fixing software single-step.

Are there [get_kernel_info] and [get_software_singlestep] testsuite calls?

Andrew