On 07/25/2018 09:17 PM, Tom Tromey wrote:
>>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:
> 
> Tom> The patch fixes the segfault by returning an error in info_program_command
> Tom> if get_last_target_status returns minus_one_ptid.
> 
> Tom> The test-case is non-standard, because the standard approach runs into
> Tom> PR23020, a problem with gdb going to the background.
> 
> Is it possible to fix this part first?

I've investigated a bit, but I think I need serious time to be able to
fix this, I'm not familiar with the tty interface.

> Otherwise it seems to me that the test case has some issues.
> 
> Also, is the bug dependent on -batch?

No, that's the way to run the test-case without running into PR23368
(gdb going to the background).

>  This wasn't clear to me, but I
> guess if it is then the approach to the test case makes more sense.
> 
> Tom> 	PR breakpoints/23366
> Tom> 	* infcmd.c (info_program_command): Handle ptid == minus_one_ptid.
> 
> This part looks ok to me.
> 
> Tom> +if { ![file exists /bin/bash] } {
> ...
> Tom> +if { ![file exists /bin/ls] } {
> ...
> Tom> +    catch {exec /bin/bash -c "$GDB $FLAGS"} catchlog
> 

I'm now using "remote_file target exists".

> I think this sort of thing can't be done when doing remote host testing,
> so some sort of check is needed for that.  I don't remember any more
> exactly how this is done but there should be other examples in the tree.
> 
> Tom> +if { ![file exists /bin/bash] } {
> Tom> +    unsupported "no bash"
> Tom> +}
> 
> Also there has to be a "return" after the "unsupported"; unsupported
> just logs a message and without the return, the rest of the test will
> still be run.
> 

Updated accordingly and committed as attached.

Thanks,
- Tom