Re: [PATCH] [COMMITTED] Fix PR gdb/17820

[Patrick Palka <patrick@parcs.ath.cx>]

On Fri, May 15, 2015 at 1:48 PM, Pedro Alves <palves@redhat.com> wrote:
> On 05/15/2015 06:26 PM, Patrick Palka wrote:
>> On Fri, May 15, 2015 at 1:17 PM, Pedro Alves <palves@redhat.com> wrote:
>>> On 05/15/2015 06:09 PM, Patrick Palka wrote:
>>>> On Fri, May 15, 2015 at 12:05 PM, Pedro Alves <palves@redhat.com> wrote:
>>>>> Hi Patrick,
>>>>>
>>>>> I noticed that the buildbots are showing that this new test is failing:
>>>>>
>>>>>  https://sourceware.org/ml/gdb-testers/2015-q2/msg04164.html
>>>>>
>>>>> ~~~
>>>>> ============================
>>>>> new FAIL: gdb.base/gdbinit-history.exp: show commands <<2>>
>>>>> new FAIL: gdb.base/gdbinit-history.exp: show history size
>>>>> new FAIL: gdb.base/gdbinit-history.exp: show history size <<2>>
>>>>
>>>> Also the tests in this file have duplicate names.. That's undesirable
>>>> right?  If so I could make the names unique.
>>>
>>> Yes.  I should have spotted that earlier.
>>>
>>>> Would such a change fall under the "obvious" rule?
>>>
>>> Not sure, depends on how you would fix it. :-)  Apply the test
>>> described in MAINTAINERS.  :-)
>>
>> So it will probably not be obvious because of naming preferences.
>
> Sorry for the trigger-happy pun.  I didn't mean to sound
> rude or put you off.  I certainly do not hate your work.  :-)

No problem, I had not inferred any rude intentions.

>
> I was just thinking that someone not familiar with the
> testsuite's history might consider obvious to change the test
> names one by one, while we avoid that nowadays in some cases
> (like described below).

Yeah, I was planning on exactly that!

>
>>
>>>
>>> There are a couple ways to address that.  In cases like
>>> this test, where we have a function that called multiple
>>> times, the modern way is to use with_test_prefix to wrap the
>>> function call or the function body, which then also covers
>>> FAILs issued from within gdb_start, etc.
>>
>> Cool.. I will do something like that.
>
> Excellent, thanks.
>
> --
> Pedro Alves
>