Re: [PATCH 1/3] gdb/testsuite: Fix broken regexp in gdbstub case

[Andrew Burgess <andrew.burgess@embecosm.com>]

* Pedro Alves <palves@redhat.com> [2018-04-13 13:12:30 +0100]:

> On 04/09/2018 04:15 PM, Andrew Burgess wrote:
> > When $use_gdb_stub is true then, when we start an MI target there's a
> > regexp to match GDB's startup pattern.  Unfortunately the pattern is
> > broken, and we're also missing a timeout case in the match list (which
> > would have helped point out that the regexp was broken).
> 
> That seems to have been added by:
>  https://sourceware.org/ml/gdb-patches/2015-12/msg00357.html
> 
> curious how it seems to have worked back then.  I wonder whether
> it was GDB's CLI-in-MI output that changed meanwhile?
> 
> I'd have appreciated a bit more detail on what is actually broken in
> the pattern.  The git log above gives no clue, and the diff is not
> exactly easy to read.  I trust that it's now correct though.  Thus, OK.

I aim to please, so here's a revised commit message detailing what
changed.  The actual patch content hasn't changed.  I'll wait a few
days, but if I don't hear anything I'll assume your approval above
still stands and apply the patch.

Thanks,
Andrew

---

gdb/testsuite: Fix broken regexp in gdbstub case

When $use_gdb_stub is true then, when we start an MI target there's a
regexp to match GDB's startup pattern.  Unfortunately the pattern is
broken, and we're also missing a timeout case in the match list (which
would have helped point out that the regexp was broken).

The changes to the regexp are:

  1. Remove '${run_match}' prefix, the issued command doesn't include
  '${run_prefix}' so expecting '${run_match}' is wrong.

  2. Replaced '\\n' with '\\\\n' in order to match literal '\n' in
  GDBs output (that is, match a backslash followed by 'n', not a
  newline character).

  3. Replaced a '.' (matching any character) with '\.' to match a '.'
  and moved the '\.' into the correct place in the regexp.

  4. Replaced '\r\n' with '[\r\n]+' to match the end of a line.  This
  change isn't esential, but matches the other end of line patterns
  within this regexp.

Here's an example of the output that the regexp should match taken
from a testfile log, the first line is the command sent to GDB, and
the remaining lines are the response from GDB:

  jump *_start
  &"jump *_start\n"
  ~"Continuing at 0x10074.\n"
  ^running
  *running,thread-id="all"
  (gdb)

gdb/testsuite/ChangeLog:

	* lib/mi-support.exp (mi_run_cmd_full): Fix regexp and add a
	timeout.

---
 gdb/testsuite/ChangeLog          | 5 +++++
 gdb/testsuite/lib/mi-support.exp | 6 +++++-
 2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/gdb/testsuite/lib/mi-support.exp b/gdb/testsuite/lib/mi-support.exp
index 2846da74e47..851e490f4de 100644
--- a/gdb/testsuite/lib/mi-support.exp
+++ b/gdb/testsuite/lib/mi-support.exp
@@ -979,7 +979,11 @@ proc mi_run_cmd_full {use_mi_command args} {
 	send_gdb  "jump *$start\n"
 	warning "Using CLI jump command, expect run-to-main FAIL"
 	gdb_expect {
-	    -re "${run_match}&\"jump \\*${start}\\n\"\[\r\n\]+~\"Continuing at 0x\[0-9A-Fa-f\]+\\n.\"\[\r\n\]+\^running\[\r\n\]+\\*running,thread-id=\"\[^\"\]+\"\r\n${mi_gdb_prompt}" {}
+	    -re "&\"jump \\*${start}\\\\n\"\[\r\n\]+~\"Continuing at 0x\[0-9A-Fa-f\]+\.\\\\n\"\[\r\n\]+\\^running\[\r\n\]+\\*running,thread-id=\"\[^\"\]+\"\[\r\n\]+${mi_gdb_prompt}" {}
+	    timeout {
+		perror "Unable to start target"
+		return -1
+	    }
 	}
 	return 0
     }
-- 
2.14.3