This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [BuildBot] Notifications disabled for Debian-s390x-* and Fedora-ppc64*-* builders


On Fri, Dec 15, 2017 at 9:48 AM, Pedro Alves <palves@redhat.com> wrote:
> On 12/15/2017 02:34 PM, David Edelsohn wrote:
>
>> I believe that the GDB buildbots should be useful to the GDB community
>> and I appreciate your efforts to maintain it.
>>
>> But the irony in your message should not be ignored.  You implicitly
>> express that the buildbots are ignored, except possibly x86 and Try
>> Builds on x86, while you proceed to ask for additional participation.
>> This is fundamentally inconsistent.  Actions speak louder than words
>> and this demonstrates the priorities of the GDB community.
>
> I think there must be some misunderstanding here.  IIUC, the problem with
> the builders is not related at all to the quality of the PPC64 or s390
> gdb ports.  The problems are something else beyond GDB's control, they're
> network, oversubscription, etc. problems, something like that.  The builder
> machines have simply not been able to exercise their basic function stably
> enough, in turn causing problems for developers, instead of helping.  What
> can the community, as in, GDB maintainers, do, then?  I for one have no
> idea what I (just a developer) can do here.  I think that the best course
> of action is to disable reporting while investigations proceed, which I
> understood it being exactly what is being done here.  I don't think
> Sergio deserved the reaction.

This is an incorrect analysis of the problem and the status.

First, a problem with the buildbot that started in September should
not be reported and addressed in December.  That clearly shows that
the community is ignoring the buildbots (at least for targets other
than x86).

Second, the s390 VM is not under-sized and should be equivalent to any
x86 VM running a buildslave.

Third, the testsuite summaries that no one from the GDB community
monitored show that the testsuite runtime jumped from a relatively
short amount of time to over 9 hours for each run, which points to a
newly introduced problem in GDB or in the testsuite (timeouts?).

Fourth, the same s390 VM runs buildbots for other projects, such as
Python, that were not adversely affected and have not seen a huge jump
in cycle time or delays.

Again, everything points to a problem in GDB, its testsuite, or its
infrastructure, not in the buildslave.  Yes, it's ironic to announce
that annoying warning emails of problems will not be ignored while
asking for more resources and participation.

Thanks, David


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]