[PATCH] gdb/testsuite: Fix rtld-step-nodebugsym.exp

[PATCH] gdb/testsuite: Fix rtld-step-nodebugsym.exp

[Lancelot SIX]

The test case introduced in bafcc335266 (Fix stepping in rtld without
debug symbol) fails on some systems as reported by PR/29768.  This can
be seen if the system does not have debug info for the libc:

  (gdb) step^M
  Single stepping until exit from function main,^M
  which has no line number information.^M
  hello world[Inferior 1 (process 48203) exited normally]^M
  (gdb) PASS: gdb.base/rtld-step-nodebugsym.exp: step
  continue^M
  The program is not being run.^M
  (gdb) FAIL: gdb.base/rtld-step-nodebugsym.exp: continue until exit (the program is no longer running)

Without glibc debug info, GDB steps until the program finishes, and
then "gdb_continue_to_end" fails.

As this test was designed to check that GDB does not crash in the "step"
command, the continue does not carry real meaning to the test.

Replace it by "print 0" so we still check that after the step command
GDB is still alive, which is what we care about.

Tested on Ubuntu-22.04 x86_64, with and without libc6-dbg.

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29768
---
 gdb/testsuite/gdb.base/rtld-step-nodebugsym.exp | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/gdb/testsuite/gdb.base/rtld-step-nodebugsym.exp b/gdb/testsuite/gdb.base/rtld-step-nodebugsym.exp
index 1c8fa478cb8..80ff1572c1a 100644
--- a/gdb/testsuite/gdb.base/rtld-step-nodebugsym.exp
+++ b/gdb/testsuite/gdb.base/rtld-step-nodebugsym.exp
@@ -29,4 +29,6 @@ if { ![runto_main] } {
 
 gdb_test "step" "Single stepping until exit from function.*"
 
-gdb_continue_to_end "" continue 1
+# GDB used to crash while doing the step command.  Check that this bug is
+# fixed and that GDB is still alive at this point.
+gdb_test "print 0" "= 0"

base-commit: bafcc335266708d5fd62159df90d139d00666ca8
-- 
2.34.1

Re: [PATCH] gdb/testsuite: Fix rtld-step-nodebugsym.exp

[Simon Marchi]

On 11/9/22 18:29, Lancelot SIX via Gdb-patches wrote:
> The test case introduced in bafcc335266 (Fix stepping in rtld without
> debug symbol) fails on some systems as reported by PR/29768.  This can
> be seen if the system does not have debug info for the libc:
> 
>   (gdb) step^M
>   Single stepping until exit from function main,^M
>   which has no line number information.^M
>   hello world[Inferior 1 (process 48203) exited normally]^M
>   (gdb) PASS: gdb.base/rtld-step-nodebugsym.exp: step
>   continue^M
>   The program is not being run.^M
>   (gdb) FAIL: gdb.base/rtld-step-nodebugsym.exp: continue until exit (the program is no longer running)
> 
> Without glibc debug info, GDB steps until the program finishes, and
> then "gdb_continue_to_end" fails.
> 
> As this test was designed to check that GDB does not crash in the "step"
> command, the continue does not carry real meaning to the test.
> 
> Replace it by "print 0" so we still check that after the step command
> GDB is still alive, which is what we care about.
> 
> Tested on Ubuntu-22.04 x86_64, with and without libc6-dbg.
> 
> Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29768
> ---
>  gdb/testsuite/gdb.base/rtld-step-nodebugsym.exp | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/gdb/testsuite/gdb.base/rtld-step-nodebugsym.exp b/gdb/testsuite/gdb.base/rtld-step-nodebugsym.exp
> index 1c8fa478cb8..80ff1572c1a 100644
> --- a/gdb/testsuite/gdb.base/rtld-step-nodebugsym.exp
> +++ b/gdb/testsuite/gdb.base/rtld-step-nodebugsym.exp
> @@ -29,4 +29,6 @@ if { ![runto_main] } {
>  
>  gdb_test "step" "Single stepping until exit from function.*"
>  
> -gdb_continue_to_end "" continue 1
> +# GDB used to crash while doing the step command.  Check that this bug is
> +# fixed and that GDB is still alive at this point.
> +gdb_test "print 0" "= 0"

In theory, the gdb_test for step expects to see a (gdb) prompt at the
end, and I suppose that GDB would crash before printing that prompt.  So
the test would catch the problem even without that "print 0".  But I
don't mind the extra safety:

Approved-By: Simon Marchi <simon.marchi@efficios.com>

Simon

Re: [PATCH] gdb/testsuite: Fix rtld-step-nodebugsym.exp

[Lancelot SIX]

> 
> In theory, the gdb_test for step expects to see a (gdb) prompt at the
> end, and I suppose that GDB would crash before printing that prompt.  So
> the test would catch the problem even without that "print 0".  But I
> don't mind the extra safety:

Hi,

Thanks for spotting the issue and for the review.

If GDB crashes, we always have an unresolved test case reported in the 
testsuite summary.  However, with this extra "p 0", the output during 
the test run is slightly more verbose as runtest complains it cannot 
issue the command, which I find slightly easier to spot.

I'll push this fix shortly.

Best,
Lancelot