[Piglit] [Patch v4 1/4] framework/dmesg: Adds a new module for dmesg

[Ilia Mirkin]

On Tue, Feb 4, 2014 at 11:07 AM, Dylan Baker <baker.dylan.c at gmail.com> wrote:
> On Tuesday, February 04, 2014 10:59:20 AM Ilia Mirkin wrote:
>> On Tue, Feb 4, 2014 at 10:50 AM, Dylan Baker <baker.dylan.c at gmail.com>
>> > +        if not self._last_message:
>> > +            # We need to ensure that there is something in dmesg to check
>> > +            # timestamps.
>> > +            raise DmesgError(
>> > +                "Your dmesg ringbuffer is empty, so it's not possible to
>> > check " +                "for timestamp support. Try adding something to
>> > dmesg by " +                "writing /dev/kmsg. ex `echo 'foo' >>
>> > /dev/kmsg")
>> TBH, I'm not a fan of that -- doing "dmesg -c" isn't such an uncommon
>> thing. Printing out a warning is fine (and then erroring out when
>> things don't pan out), but I dunno about making it a hard requirement
>> to have something in dmesg.
>>
>
> Yeah, I'm not super happy with it either. I just don't see anywhere in the
> actually called methods we can bail out without having unexpected results
> first, without doing something as awful as a check for timestamps in
> dmesg_update(), which would lead to a bad performance hit. Any suggestions?

Print a warning and let things fail as they go? I prefer to allow
people to shoot themselves in the foot. Also what will _actually_
fail? The timestamp isn't parsed anywhere, the only thing that'll go
wrong is that some dmesg errors might not get picked up if they're
overly repetitive.

  -ilia