Re: [PATCH 2/3] [gdb/testsuite] Use gdb_exit_cleanup a bit more

[Tom de Vries <tdevries@suse.de>]

On 9/27/26 10:52 AM, Andrew Burgess wrote:
> Tom de Vries <tdevries@suse.de> writes:
> 
>> I ran into trouble running test-case gdb.base/quit-live.exp using a remote
>> host configuration:
>> - host board local-remote-host
>> - target board remote-gdbserver-on-localhost.
>>
>> The problem is that the test-case makes gdb quit without updating
>> gdb_spawn_id.  Consequently, default_gdb_exit tries to exit gdb.
>> It does so by sending quit to gdb and waiting for it to exit, which get us:
>> ...
>> ERROR: : spawn id exp9 not open
>> ...
>>
>> Fix this using gdb_exit_cleanup.  Likewise in gdb.threads/killed.exp.
>> ---
>>   gdb/testsuite/gdb.base/quit-live.exp | 1 +
>>   gdb/testsuite/gdb.threads/killed.exp | 1 +
>>   2 files changed, 2 insertions(+)
>>
>> diff --git a/gdb/testsuite/gdb.base/quit-live.exp b/gdb/testsuite/gdb.base/quit-live.exp
>> index 14f87f9b7f8..8dae4fe2aca 100644
>> --- a/gdb/testsuite/gdb.base/quit-live.exp
>> +++ b/gdb/testsuite/gdb.base/quit-live.exp
>> @@ -152,6 +152,7 @@ proc quit_with_live_inferior {appear_how extra_inferior quit_how} {
>>   		gdb_test_multiple "" $test {
>>   		    eof {
>>   			pass $test
>> +			gdb_exit_cleanup
> 
> There's another place in this test script which detects eof from GDB,
> but you've not added the gdb_exit_cleanup call there.
> 
> Should we also be patching that location?  If not why not?  And if the
> answer is there's a good reason why not, then I think it is worth
> mentioning in the commit message, and as a comment at that location in
> the code.
> 

Hi Andrew,

thanks for the review(s).

The focus of this commit is to address a specific error, and the 
addition in quit_with_live_inferior fixes it.

I did not encounter the same error in test_quit_with_sig, but it's 
probably a good idea to apply the same pattern.

[ It might even be necessary on msys2, but AFAIR currently the test-case 
fails in such a way that the eof is not reached.  I briefly tried making 
the testcase work using some kill equivalent, but that didn't work out. ]

Anyway, I've added the gdb_exit_cleanup in test_quit_with_sig, and I've 
updated the commit message to:
...
     Fix this in quit_with_live_inferior using gdb_exit_cleanup.
     Likewise in gdb.threads/killed.exp.

     While we're at it, also add default_gdb_exit for another eof clause
     in gdb.base/quit-live.exp, in proc test_quit_with_sig.
...
and pushed.

Thanks,
- Tom

> Thanks,
> Andrew
>