Re: [PATCH 2/2,v3] [AArch64] Test handling of additional brk instruction patterns

[Simon Marchi <simark@simark.ca>]

On 2020-01-29 6:30 a.m., Luis Machado wrote:
>>> +# Number of expected SIGTRAP's to get.  This needs to be kept in sync
>>> +# with the source file.
>>> +set expected_traps 3
>>> +set keep_going 1
>>> +set count 0
>>> +set old_timeout $timeout
>>> +set timeout 10
>>
>> Any reason you are changing the timeout?  There is nothing in the test that
>> looks like it would take time.
>>
> 
> If GDB doesn't support one of these instructions, it will be caught in 
> an infinite loop. The reduced timeout will prevent a long wait time 
> until we bail out.

Ok.  The worry I have with it is that if the target board has raised the timeout
on purpose, because it's testing with a slow target/link/emulator, then this will
cancel it.  In this case I'd just leave the timeout as it is.  Normally, GDB won't
be broken for this test case, so it won't matter.

Moreover, the timeout here on my x86 machine is 10 seconds by default.  I just checked
on an AArch64 box on the compile farm, it's 10 there too.  So in which case was it
useful to set it to 10?

>> If changing the timeout is really necessary, look into using with_timeout_factor.
>>
> 
> That would raise the timeout even further. We want a reduced one.
> 
> It would be nice if we could reduce the timeout with 
> with_timeout_factor. I gave it a try but it didn't work.
> 
> I think we need adjustments to make it work with a floating point 
> number. I'll look into it.

I'm not sure that's desirable for a test case to ever reduce the timeout, for the reason
explained above.

Simon