[review v2] testsuite: add test for handling multiple inferiors exiting

[Pedro Alves (Code Review)]

Pedro Alves has posted comments on this change.

Change URL: https://gnutoolchain-gerrit.osci.io/r/c/binutils-gdb/+/134
......................................................................


Patch Set 2:

(6 comments)

| --- /dev/null
| +++ /COMMIT_MSG
| @@ -1,0 +7,31 @@ CommitDate: 2019-11-04 11:21:38 +0100
| +testsuite: add test for handling multiple inferiors exiting
| +
| +In all-stop mode, upon receiving a stop event, GDB attempts to stop
| +other threads.  It is possible that a thread GDB wants to stop has
| +already terminated before GDB sees that exit event.  So, the response
| +to GDB's stop request may be a status kind *other* than
| +TARGET_WAITKIND_STOPPED (e.g.  TARGET_WAITKIND_EXITED).  Such a case
| +should be handled by GDB gracefully and the user should be given back
| +the debugger prompt.
| +
| +The new tests added with this patch cover the TARGET_WAITKIND_EXITED
| +and TARGET_WAITKIND_SIGNALLED cases.  The existing tests
| +
| +  gdb.threads/leader-exit.exp
| +  gdb.threads/non-ldr-exc-2.exp
| +
| +exercise the TARGET_WAITKIND_NO_RESUMED response, whereas the existing
| +test
| +
| +  gdb.threads/tid-reuse.exp
| +
| +exercises the TARGET_WAITKIND_THREAD_EXITED case.

PS2, Line 28:

I'm curious about these references to the other waitkind cases.  I'm
not sure I understand correctly what they mean though, because those
tests already pass today without the fix you proposed?  The fix
included a change to the TARGET_WAITKIND_THREAD_EXITED case, for
example.

| +
| +gdb/testsuite/ChangeLog:
| +2019-11-04  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>
| +
| +	* gdb.multi/multi-exit.c: New file.
| +	* gdb.multi/multi-exit.exp: New file.
| +	* gdb.multi/multi-kill.c: New file.
| +	* gdb.multi/multi-kill.exp: New file.
| +
| --- /dev/null
| +++ gdb/testsuite/gdb.multi/multi-exit.exp
| @@ -1,0 +25,20 @@ # TARGET_WAITKIND_STOPPED should be handled by GDB without problems.
| +standard_testfile
| +
| +if {![can_spawn_for_attach]} {
| +    return 0
| +}
| +
| +if {[prepare_for_testing "failed to prepare" $testfile $srcfile]} {
| +    return -1
| +}
| +
| +gdb_test "add-inferior -exec $binfile"

PS2, Line 35:

Please add an explicit test name, so that the local path in $binfile
doesn't show up in gdb.sum.

| +
| +proc test {} {
| +    # Start the processes separately.
| +    gdb_test_no_output "set schedule-multiple off"
| +
| +    gdb_test "inferior 1" ".*Switching to inferior 1.*"
| +    if {![runto_main]} {
| +	fail "starting inferior 1"
| +	return -1
| --- /dev/null
| +++ gdb/testsuite/gdb.multi/multi-kill.exp
| @@ -1,0 +26,64 @@ standard_testfile
| +
| +if {![can_spawn_for_attach]} {
| +    return 0
| +}
| +
| +if {[prepare_for_testing "failed to prepare" $testfile $srcfile]} {
| +    return -1
| +}
| +
| +gdb_test "add-inferior -exec $binfile" "Added inferior 2.*"

PS2, Line 35:

Same comment about paths in gdb.sum.

| +
| +# We want both processes in a running state.
| +gdb_test_no_output "set schedule-multiple on"
| +
| +proc test {} {
| +    # Start the programs, attach to them, then kill both from outside.
| +    global binfile
| +    global gdb_prompt
| +
| +    set spawn_id_list [spawn_wait_for_attach [list $binfile $binfile]]
| +    set test_spawn_id1 [lindex $spawn_id_list 0]
| +    set test_spawn_id2 [lindex $spawn_id_list 1]
| +    set testpid1 [spawn_id_get_pid $test_spawn_id1]
| +    set testpid2 [spawn_id_get_pid $test_spawn_id2]
| +
| +    gdb_test "inferior 1" ".*Switching to inferior 1.*"
| +
| +    gdb_test "attach $testpid1" \
| +	"Attaching to program: .*, process $testpid1.*(in|at).*" \
| +	"attach to program 1"

PS2, Line 55:

When testing with --target_board=native-extended-remote, I get:

 attach 22448
 A program is being debugged already.  Kill it? (y or n) n
 Not killed.
 (gdb) FAIL: gdb.multi/multi-kill.exp: iteration 2: attach to program 1 (got interactive prompt)

| +
| +    gdb_test "inferior 2" ".*Switching to inferior 2.*"
| +
| +    gdb_test "attach $testpid2" \
| +	"Attaching to program: .*, process $testpid2.*(in|at).*" \
| +	"attach to program 2"
| +
| +    send_gdb "continue\n"
| +    gdb_expect {

PS2, Line 64:

Please use gdb_test_multiple instead of gdb_expect.  Instead of
"default" use the proc's return value.

| +	-re "Continuing." {
| +	    # nothing; keep going.
| +	}
| +	default {
| +	    untested "could not continue"
| +	    return -1
| +	}
| +    }
| +
| +    # Kill both processes at once.
| +    remote_exec build "kill -9 ${testpid1} ${testpid2}"
| +
| +    # GDB should give the prompt back instead of going into an
| +    # infinite loop.
| +    set testname "back to gdb prompt"
| +    gdb_expect {

PS2, Line 80:

Ditto.

| +	-re "Program terminated with signal.*$gdb_prompt" {
| +	    pass $testname
| +	}
| +	default {
| +	    fail $testname
| +	}
| +    }
| +
| +    # Make sure that the processes are gone.

-- 
Gerrit-Project: binutils-gdb
Gerrit-Branch: master
Gerrit-Change-Id: Ib6fe193d23ba08d7fbb369f41debf2cee5816f6b
Gerrit-Change-Number: 134
Gerrit-PatchSet: 2
Gerrit-Owner: Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
Gerrit-Reviewer: Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
Gerrit-CC: Luis Machado <luis.machado@linaro.org>
Gerrit-CC: Pedro Alves <palves@redhat.com>
Gerrit-Comment-Date: Fri, 06 Dec 2019 18:21:10 +0000
Gerrit-HasComments: Yes
Gerrit-Has-Labels: No
Gerrit-MessageType: comment