Re: [PATCH 5/8] gen-pert-test: parallel build support

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

On Sat, Jul 25, 2015 at 2:25 PM, Doug Evans <xdje42@gmail.com> wrote:
> On Sat, Jul 25, 2015 at 9:44 AM, Patrick Palka <patrick@parcs.ath.cx> wrote:
>> On Tue, Jul 21, 2015 at 9:44 AM, Doug Evans <dje@google.com> wrote:
>>> Hi.
>>>
>>> This patch adds parallel build support for perf testcases.
>>> To use the existing machinery, GDB_PARALLEL now contains
>>> the subdirectory in which to put the parallel builds.
>>>
>>> 2015-07-20  Doug Evans  <dje@google.com>
>>>
>>>         * Makefile.in (check/%.exp): Pass directory for GDB_PARALLEL.
>>>         (workers/%.worker, build-perf): New rule.
>>>         (GDB_PERFTEST_MODE): New variable.
>>>         (check-perf): Use it.
>>>         (clean): Clean up gdb.perf parallel build subdirs.
>>>         * lib/build-piece.exp: New file.
>>>         * lib/cache.exp (gdb_do_cache): Include $GDB_PARALLEL in path name.
>>>         * lib/gdb.exp (standard_output_file): Include $GDB_PARALLEL in path
>>>         name.
>>>         (standard_temp_file): Ditto.
>>>         (GDB_PARALLEL handling): Make outputs,temp,cache directories as
>>> subdirs
>>>         of $GDB_PARALLEL.
>>
>> This patch seems to have caused a number of regressions:
>>
>> http://gdb-build.sergiodj.net/builders/Fedora-x86_64-m64/builds/1502
>> http://gdb-build.sergiodj.net/builders/Fedora-x86_64-m32/builds/1501
>> http://gdb-build.sergiodj.net/builders/Fedora-i686/builds/1510
>>
>> (possibly among others)
>>
>> Reverting this patch locally makes the regressions disappear for me,
>> at least on x86_64-m64.
>
> I can't repro this.
> It's odd that this patch would cause these particular regressions:
>
> http://gdb-build.sergiodj.net/builders/Fedora-x86_64-m32/builds/1501/steps/regressions/logs/regressions
>
> http://gdb-build.sergiodj.net/builders/Fedora-x86_64-m64/builds/1502/steps/regressions/logs/regressions
>
> Also, I don't understand this one:
>
> http://gdb-build.sergiodj.net/builders/Debian-x86_64-m64/builds/1420
>
> I can imagine makefile hacking for the perf tests breaking normal make check,
> but then I'd expect the damage to be far more extensive, but the patch
> reported in 1420 cannot have caused those regressions.

In this case, the regressions listed for build 1420 originated from build 1417:

http://gdb-build.sergiodj.net/builders/Debian-x86_64-m64/builds/1417

The baseline for checking for regressions is not the build before, but
rather it's a "rolling" baseline or something like that.