llvm dev - Aug 2017 - [RFC] 'Review corner' section in LLVM Weekly

On Mon, Aug 28, 2017 at 8:04 AM, Alex Bradbury <asb at asbradbury.org>
wrote:
> On 27 August 2017 at 00:01, Alex Bradbury <asb at asbradbury.org>
wrote:
>> Hi all. I'm assuming most people reading this email are familiar
with LLVM's
>> code review process
<http://llvm.org/docs/DeveloperPolicy.html#code-reviews>
>> as well as LLVM Weekly, the development newsletter I've written and
sent out
>> every Monday since Jan 2014. Since that time, it's provided
something of a
>> "signal boost" for important mailing list discussions and
commits. I feel it
>> could play a similar role in helping patches that are stuck waiting for
code
>> reviews, or drawing attention to submissions from first time
contributors.
>> There may be alternative or complementary approaches to tackling this
>> perceived problem we should discuss - I'm coming from a position of
trying to
>> apply the tools I have at my disposal. Also see my previous thoughts on
this
>> issue
<http://lists.llvm.org/pipermail/llvm-dev/2016-November/106696.html>.
>
> Hans Wennborg suggested on Twitter that bugs could also be included. I
> can't write a coherent response in 140 characters, so am responding
> here.
The reason I brought it up is because I think the situation is
somewhat similar to stalled reviews; in fact forgotten bug reports are
even worse in a way: if no developers are cc'd on the bug, the pings
don't go anywhere -- no mailing list, no nothing. At least with code
reviews, pings usually go to one of the -commits lists.
> I think that highlighting important bugs is also a useful activity,
> but I have less faith that there's something LLVM Weekly can do to
> help here. There may be some value in highlighting bugs which are
> release blockers (though I already try to make sure I link to any such
> list when posted by the release manager) or a selection of
'beginner'
> bugs, (props to Brian Gesiak for pushing for this category). But
> beyond that, how do you decide which bugs to highlight? With patches,
> it's normally the case that a highly motivated party (the patch
> author) has put in the majority of the work, and a relatively smaller
> amount of incremental effort is required from others in the LLVM
> community (code review). With bugs it's the other way around - a huge
> amount of additional work might be required to properly diagnose and
> address a bug report.
>
> I'm totally open to trying something if you think there's a way
LLVM
> Weekly can have an impact in this area, but I'm less hopeful about the
> potential impact in reducing the number of open bugs.
What I was thinking when I replied to your tweet was something like
"bugs filed in the last 7 days which no-one seems to have looked at",
or something similar. Hopefully it should be possible to build the
list automatically.

I'm not sure how large that list would be each week though, or how
useful it would be. Also, this might be something we should fix by
having a better bug triaging process in general.

Cheers,
Hans

On 28 August 2017 at 17:49, Hans Wennborg <hans at chromium.org>
wrote:
> On Mon, Aug 28, 2017 at 8:04 AM, Alex Bradbury <asb at
asbradbury.org> wrote:
>> On 27 August 2017 at 00:01, Alex Bradbury <asb at asbradbury.org>
wrote:
>>> Hi all. I'm assuming most people reading this email are
familiar with LLVM's
>>> code review process
<http://llvm.org/docs/DeveloperPolicy.html#code-reviews>
>>> as well as LLVM Weekly, the development newsletter I've written
and sent out
>>> every Monday since Jan 2014. Since that time, it's provided
something of a
>>> "signal boost" for important mailing list discussions and
commits. I feel it
>>> could play a similar role in helping patches that are stuck waiting
for code
>>> reviews, or drawing attention to submissions from first time
contributors.
>>> There may be alternative or complementary approaches to tackling
this
>>> perceived problem we should discuss - I'm coming from a
position of trying to
>>> apply the tools I have at my disposal. Also see my previous
thoughts on this
>>> issue
<http://lists.llvm.org/pipermail/llvm-dev/2016-November/106696.html>.
>>
>> Hans Wennborg suggested on Twitter that bugs could also be included. I
>> can't write a coherent response in 140 characters, so am responding
>> here.
>
> The reason I brought it up is because I think the situation is
> somewhat similar to stalled reviews; in fact forgotten bug reports are
> even worse in a way: if no developers are cc'd on the bug, the pings
> don't go anywhere -- no mailing list, no nothing. At least with code
> reviews, pings usually go to one of the -commits lists.
Good point.
> What I was thinking when I replied to your tweet was something like
> "bugs filed in the last 7 days which no-one seems to have looked
at",
> or something similar. Hopefully it should be possible to build the
> list automatically.
>
> I'm not sure how large that list would be each week though, or how
> useful it would be. Also, this might be something we should fix by
> having a better bug triaging process in general.
Thanks for the clarification. Trying to improve bug triage problem
seems like a manageable proposition. I'd guess it's already possible
to construct a bugzilla query showing bugs filed in the past 7 days
that haven't been triaged?

Something like the following?:

"""
## Bug corner

In the past 7 days, XXX bugs have been opened, YYY bugs have been
closed. ZZZ newly-filed bugs remain untriaged - see <link> to help.
Thanks to contributions from CREDIT1, ..., CREDITN, AAA out of BBB
bugs were triaged last week.
"""

Best,

Alex

> -----Original Message-----
> From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of
Alex
> Bradbury via llvm-dev
> Sent: Monday, August 28, 2017 10:02 AM
> To: Hans Wennborg
> Cc: llvm-dev
> Subject: Re: [llvm-dev] [RFC] 'Review corner' section in LLVM
Weekly
> 
> On 28 August 2017 at 17:49, Hans Wennborg <hans at chromium.org>
wrote:
> > On Mon, Aug 28, 2017 at 8:04 AM, Alex Bradbury <asb at
asbradbury.org>
> wrote:
> > What I was thinking when I replied to your tweet was something like
> > "bugs filed in the last 7 days which no-one seems to have looked
at",
> > or something similar. Hopefully it should be possible to build the
> > list automatically.
> >
> > I'm not sure how large that list would be each week though, or how
> > useful it would be. Also, this might be something we should fix by
> > having a better bug triaging process in general.
> 
> Thanks for the clarification. Trying to improve bug triage problem
> seems like a manageable proposition. I'd guess it's already
possible
> to construct a bugzilla query showing bugs filed in the past 7 days
> that haven't been triaged?
> 
> Something like the following?:
> 
> """
> ## Bug corner
> 
> In the past 7 days, XXX bugs have been opened, YYY bugs have been
> closed. ZZZ newly-filed bugs remain untriaged - see <link> to help.
> Thanks to contributions from CREDIT1, ..., CREDITN, AAA out of BBB
> bugs were triaged last week.
> """
As I reported back in June, the overall LLVM open bug count increases
by an average of 4 per day.  I don't know how many of the bugs that
stay open get comments, but this gives you a reasonable upper bound
on the size of this kind of weekly summary.
--paulr