Re: [PATCH v2 0/8] Remove XML files from gdbserver

[Alan Hayward <Alan dot Hayward at arm dot com>]

> On 29 Jan 2018, at 17:13, Alan Hayward <Alan.Hayward@arm.com> wrote:
> 
> Only just spotted this thread.
> 
>> On 29 Jan 2018, at 16:28, Yao Qi <qiyaoltc@gmail.com> wrote:
>> 
>> Hi Omair,
>> Thanks for testing these patches...
>> 
>> On Fri, Jan 26, 2018 at 10:31 PM, Omair Javaid <omair.javaid@linaro.org> wrote:
>>> Hi Alan,
>>> 
>>> I just ran gdb testsuite with native-gdbserver configuration on your
>>> patches and found a couple of regressions. At least first three seems to be
>>> coming out of your patch series as they seem persistent over multiple runs
>>> of testsuite.
>>> 
>>> 1c1
>>> < Test Run By omair on Sat Jan 27 02:51:24 2018
>>> ---
>>>> Test Run By omair on Sat Jan 27 01:58:06 2018
>>> 1257,1258c1257,1258
>>> < FAIL: gdb.arch/i386-mpx.exp: bndcfgu formating
>>> < FAIL: gdb.arch/i386-mpx.exp: test if bndstatus is enabled
>>> ---
>>>> PASS: gdb.arch/i386-mpx.exp: bndcfgu formating
>>>> PASS: gdb.arch/i386-mpx.exp: test if bndstatus is enabled
>>> 6706c6706
>>> < FAIL: gdb.base/gcore.exp: corefile restored all registers
>>> ---
>>>> PASS: gdb.base/gcore.exp: corefile restored all registers
>>> 50790c50790
>> 
>> If I read the diff correctly, there are regressions in
>> gdb.arch/i386-mpx.exp and gdb.base/gcore.exp.
> 
> Looking into it, I always get UNTESTED: gdb.arch/i386-mpx.exp: failed to prepare
> 
> MPX was a Skylake addition. I’ve been using a Sandy Bridge processor.
> 
> Is there any emulator setup or suchlike to enable testing?
> 
> I’ll see if I can find a suitable box to use.

Was finally able to run the mpx tests.
(My MacBook is a Skylake, but is running macOS. Using ubuntu in virtualbox disables
mpx, but using ubuntu in vmware fusion is fine! Thankfully didn’t have to resort to a
live usb stick)

On the latest head, with or without my patches I always get the same results:

Mpx tests pass with unix / no target board, but fail with native-gdbserver

"gcore.exp: corefile restored all registers” fails with unix / no target board, and passes
with native-gdbserver.

I can’t see any obvious reason why my patches would break mpx or the gcore tests,
but never say never.

Is it worth me digging more into these tests?

> 
> 
>> 
>>> < PASS: gdb.threads/thread-unwindonsignal.exp: continue until exit
>>> ---
>>>> FAIL: gdb.threads/thread-unwindonsignal.exp: continue until exit (timeout)
>>> 55145,55146c55145,55146
>> 
>> This test case isn't stable for me, I remember I saw the fail several times
>> before.
>> 
>> -- 
>> Yao (齐尧)