bugzilla voting

bugzilla voting

[Matt Hargett]

Hi,

I just wanted to express my desire for voting to be turned on in 
bugzilla. I saw the previous discussion where it was decided to leave it 
off until regressions were fixed. It makes sense for regressions to be 
the priority right now, of course, but I don't see how turning voting on 
will confuse that.

I am making this request because there are some serious code generation 
issues with the sh4 target that were reported some time ago and still do 
not appear to be fixed. I think if the user community were allowed to 
easily express their interest in certain bugs being fixed that these 
kinds of things might not appear to be so easily forgotten.

No flames, please, I just wanted to make my case for the suggestion :)

Re: bugzilla voting

[Joe Buck]

On Thu, Apr 01, 2004 at 12:56:53PM -0800, Matt Hargett wrote:
> I just wanted to express my desire for voting to be turned on in 
> bugzilla....
 
> I am making this request because there are some serious code generation 
> issues with the sh4 target that were reported some time ago and still do 
> not appear to be fixed.

Bugs are not fixed because someone votes that they be fixed, and this would
not be changed if the bugzilla voting feature is enabled.  They are fixed
because someone contributes a fix.

If you are not capable of producing a fix yourself, and you work for a
company that has a financial interest in improvements to the sh4 target,
hiring a consultant is a possibility.

But if you think that Bugzilla voting is going to speed up bug fixes to
an embedded target, I'm afraid that you would be disappointed.  Even if
voting did affect the priority assigned to bugs, people interested on more
widely used platforms would outvote you.

Re: bugzilla voting

[Tom Tromey]

>>>>> "Joe" == Joe Buck <Joe.Buck@synopsys.COM> writes:

Joe> But if you think that Bugzilla voting is going to speed up bug fixes to
Joe> an embedded target, I'm afraid that you would be disappointed.  Even if
Joe> voting did affect the priority assigned to bugs, people interested on more
Joe> widely used platforms would outvote you.

I think we could turn it on as long as everybody is aware that the
votes are non-binding.  Occasionally someone will go to bugzilla to
see what needs to be fixed; information about what other people think
is most important can be useful for prioritizing.  I'd find this
information interesting for libgcj bugs, at least when I'm working on
that.

Tom

Re: bugzilla voting

[Jamie Lokier]

Joe Buck wrote:
> But if you think that Bugzilla voting is going to speed up bug fixes to
> an embedded target, I'm afraid that you would be disappointed.  Even if
> voting did affect the priority assigned to bugs, people interested on more
> widely used platforms would outvote you.

People working on an embedded target could look at the votes
constrained to bugs which affect only that target.

While that doesn't automatically cause popular bugs to be fixed, it
does gives developers a little more information.  Sometimes that
information can even be used to justify financing work, e.g. from a cpu
manufacturer who has interest in the quality of their port.

-- Jamie

Re: bugzilla voting

[Joe Buck]

On Fri, Apr 02, 2004 at 01:32:04AM +0100, Jamie Lokier wrote:
> Joe Buck wrote:
> > But if you think that Bugzilla voting is going to speed up bug fixes to
> > an embedded target, I'm afraid that you would be disappointed.  Even if
> > voting did affect the priority assigned to bugs, people interested on more
> > widely used platforms would outvote you.
> 
> People working on an embedded target could look at the votes
> constrained to bugs which affect only that target.
> 
> While that doesn't automatically cause popular bugs to be fixed, it
> does gives developers a little more information.  Sometimes that
> information can even be used to justify financing work, e.g. from a cpu
> manufacturer who has interest in the quality of their port.

If the GCC developers think turning on voting would be useful to them, I
would be all for it.  Whether users think it's a good idea, I think,
should not be relevant.

If it is turned on, we need to be sure that expectations are set correctly
("Mommy, they fixed Johnny's bug, even though Billy had more votes!
Unfair!  I'm telling RMS!").

Re: bugzilla voting

[Jamie Lokier]

Joe Buck wrote:
> If the GCC developers think turning on voting would be useful to them, I
> would be all for it.  Whether users think it's a good idea, I think,
> should not be relevant.

I agree.

> If it is turned on, we need to be sure that expectations are set correctly
> ("Mommy, they fixed Johnny's bug, even though Billy had more votes!
> Unfair!  I'm telling RMS!").

A short text on the vote form and/or results to that effect?

"Bugs don't get fixed because they are popular: they get fixed because
someone funds a developer to do the work, or does the work themselves.
Developers generally follow the priorities of those who pay for work
to be done.  If you genuinely desire a bug to be fixed, you are
expected to fund a developer to fix it, or fix it yourself."

If you manage to attach a $1 payment option to each vote, paid in
advance, with a shortcut for submitting multiple votes, that would be
great.  Thanks! :)

-- Jamie

Re: bugzilla voting

[Robert Dewar]

Joe Buck wrote:

> If the GCC developers think turning on voting would be useful to them, I
> would be all for it.  Whether users think it's a good idea, I think,
> should not be relevant.
> 
> If it is turned on, we need to be sure that expectations are set correctly
> ("Mommy, they fixed Johnny's bug, even though Billy had more votes!
> Unfair!  I'm telling RMS!").

Since the right expectation is that votes should be entirely ignored, I
don't see any better solution to the problem of possible incorrect
expectations than to leave voting turned off!

Re: bugzilla voting

[Jamie Lokier]

Robert Dewar wrote:
> Since the right expectation is that votes should be entirely ignored, I
> don't see any better solution to the problem of possible incorrect
> expectations than to leave voting turned off!

Let them vote but hide the results! :)

-- Jamie

Re: bugzilla voting

[Per Abrahamsen]

Joe Buck <Joe.Buck@synopsys.COM> writes:

> If it is turned on, we need to be sure that expectations are set correctly
> ("Mommy, they fixed Johnny's bug, even though Billy had more votes!
> Unfair!  I'm telling RMS!").

Does this ever happen for Mozilla, or for other projects with bugzilla
voting enabled?

Re: bugzilla voting

[Per Abrahamsen]

Robert Dewar <dewar@gnat.com> writes:

> Since the right expectation is that votes should be entirely ignored,

I'd expect votes to generate:

1. A mixed level of interest among volunteers working on GCC in their
   spare time, depending on their personal motivations for doing so.

2. A low, but non-zero level of interest from people working for
   companies that does not make money directly on GCC, but see a
   strategic interest in the the existence of a good free compiler
   suite.

3. Zero interest from companies that make money directly from support
   or development of GCC.

Would that be totally off?

Re: bugzilla voting

[Richard Earnshaw]

> 3. Zero interest from companies that make money directly from support
>    or development of GCC.

Or maybe even negative interest (they'd not fix a bug with a large number 
of votes because somebody might be more likely to cough up real money for 
doing the work...)

Re: bugzilla voting

[Scott Robert Ladd]

Per Abrahamsen wrote:
> Joe Buck <Joe.Buck@synopsys.COM> writes:
>>If it is turned on, we need to be sure that expectations are set correctly
>>("Mommy, they fixed Johnny's bug, even though Billy had more votes!
>>Unfair!  I'm telling RMS!").
> 
> Does this ever happen for Mozilla, or for other projects with bugzilla
> voting enabled?

Such complainst arise often in the Java world all the time; people vote 
for bugs, and then Sun often ignore the votes in selecting what to fix 
or add. Many critical bugs (not feature requests) have high vote counts 
but lack action from Sun.

I would like to see a voting system, so long as there are reasonable 
assurances that the vote has some value. I don't expect the tyranny of 
the commons to drive every decision by the SC or release manager -- I 
would, however, want to see respect for the desires of the user community.

..Scott

-- 
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing

An Immodest Proposal for GCC Voting

[Scott Robert Ladd]

In light of the "it only gets done if it's paid for", I make the 
following proposal.

Charge for voting -- a poll tax, if you will.

That's right -- if I wanted to vote for bug #14468 (which sticks in my 
craw at the moment), I could attach a donation to said vote; the votes 
would then be weighed in importance by how much money was attached to them.

Whoever fixes the bug gets a "cut" of the attached dollars, with the 
rest going into the general GCC fund for wild SC parties and other 
administrative purposes.

We could consider it a bounty system, which has proven popular in other 
segemnts of the FOSS community.

Let's add money to voting; it's a long-time tradition!

-- 
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing

Re: bugzilla voting

[Per Abrahamsen]

Scott Robert Ladd <coyote@coyotegulch.com> writes:

> Such complainst arise often in the Java world all the time; people
> vote for bugs, and then Sun often ignore the votes in selecting what
> to fix or add. Many critical bugs (not feature requests) have high
> vote counts but lack action from Sun.

I should have specified "free software projects".  The expectations
are different for unfree software, where lobbying often is the only
way to affect the development.

> I would like to see a voting system, so long as there are reasonable
> assurances that the vote has some value.

That sentiment would be an argument against enabling voting.  I
certainly would not want to give assurances any stronger than that I
just *might* take votes into account occasionally.

Re: bugzilla voting

[Joel Sherrill]

Richard Earnshaw wrote:

>>3. Zero interest from companies that make money directly from support
>>   or development of GCC.
> 
> 
> Or maybe even negative interest (they'd not fix a bug with a large number 
> of votes because somebody might be more likely to cough up real money for 
> doing the work...)

I would expect that bugs which impact the configuratinos with the
highest numbers of users would end up getting the most votes.  So
a GNU/Linux x86 would likely generate more votes than something for
an embedded platform.

Which leads me to believe that a more accurate prioritization technique
would be to place weight on which platforms, packages, etc were
negatively impacted by the bug. So miscompiling glibc on an x86 would
be "more important" than fixing a miscompilation of any other package
on the pdp11.


-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel@OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
Support Available                (256) 722-9985

Re: bugzilla voting

[Scott Robert Ladd]

Per Abrahamsen wrote:
> I should have specified "free software projects".  The expectations
> are different for unfree software, where lobbying often is the only
> way to affect the development.

Ah, but isn't this contradicted by Mr. Dewar's argument (and yours 
below) that lobby has no value unless accompanied by money?

SRL>I would like to see a voting system, so long as there are
SRL>reasonable assurances that the vote has some value.

> That sentiment would be an argument against enabling voting.  I
> certainly would not want to give assurances any stronger than that I
> just *might* take votes into account occasionally.

If that is your attitude, and it reflects the attitude of other primary 
GCC developers, then voting or lobbying are indeed meaningless.


-- 
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing

Re: bugzilla voting

[Ian Lance Taylor]

Scott Robert Ladd <coyote@coyotegulch.com> writes:

> > That sentiment would be an argument against enabling voting.  I
> > certainly would not want to give assurances any stronger than that I
> > just *might* take votes into account occasionally.
> 
> If that is your attitude, and it reflects the attitude of other
> primary GCC developers, then voting or lobbying are indeed meaningless.

Speaking as someone who actually does fix random PRs, I think voting
would be useful.  As everyone else has said, it obviously would not be
any sort of guarantee.

Right now my random PR fixing is driven mainly by the list of PRs
reported against 3.4.0 (there are only five remaining 3.4.0 PRs in the
search which I use; right now I'm slowly poking at PR 14400).

Ian

Re: bugzilla voting

[Daniel Berlin]

>
> If you manage to attach a $1 payment option to each vote, paid in
> advance, with a shortcut for submitting multiple votes, that would be
> great.  Thanks! :)
>

I'll replace the "vote" button with a "paypal donate" button.
:)

> -- Jamie
>

Re: bugzilla voting

[Scott Robert Ladd]

Ian Lance Taylor wrote:
> Scott Robert Ladd <coyote@coyotegulch.com> writes:
> 
> 
>>>That sentiment would be an argument against enabling voting.  I
>>>certainly would not want to give assurances any stronger than that I
>>>just *might* take votes into account occasionally.
>>
>>If that is your attitude, and it reflects the attitude of other
>>primary GCC developers, then voting or lobbying are indeed meaningless.
> 
> Speaking as someone who actually does fix random PRs, I think voting
> would be useful.  As everyone else has said, it obviously would not be
> any sort of guarantee.

I much appreciate people such as yourself; I wish I had more time to 
work on GCC myself, beyond my own paltry efforts on OpenMP and gfortran.

The value of voting directly corresponds to the amount of influence it 
has -- as per the last U.S. Presidential election, where the popular 
vote was overridden by judicial fiat. If people vote for bugs, they will 
quickly become disillusioned by Per's "*might* take votes into account 
occasionally" viewpoint.

Please note that I am not criticizing Pers! A purely democratic approach 
won't work, either, as under such, the only bugs that would get fixed 
are those for very popular platforms. A system of checks and bal;ances 
would be needed to free developers from absolute dictates while also 
assuring voters that their votes mean something. Tricky.

> Right now my random PR fixing is driven mainly by the list of PRs
> reported against 3.4.0 (there are only five remaining 3.4.0 PRs in the
> search which I use; right now I'm slowly poking at PR 14400).

I'm only working in the 3.5 (tree-ssa) branch, given that OpenMP and 
gfortran reside there. I've got a couple of patches waiting to go in as 
soon as tree-ssa merges...

-- 
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing

Re: bugzilla voting

[Joe Buck]

On Fri, Apr 02, 2004 at 11:24:59AM -0500, Scott Robert Ladd wrote:
> The value of voting directly corresponds to the amount of influence it 
> has -- as per the last U.S. Presidential election, where the popular 
> vote was overridden by judicial fiat. If people vote for bugs, they will 
> quickly become disillusioned by Per's "*might* take votes into account 
> occasionally" viewpoint.

There's also the factor that "voting" on the Internet is usually dominated
by organized efforts to stuff the ballot boxes (mailing lists and weblogs
urging people to go to a site they never read and vote a certain way in some
poll).  That's why you'll see wild swings when, say, CNN puts up a poll
on a partisan issue: some popular right-wing blog might send its people there,
and then some other popular left-wing blog will do the same, etc.

> Please note that I am not criticizing Pers! A purely democratic approach 
> won't work, either, as under such, the only bugs that would get fixed 
> are those for very popular platforms.

In practice, factors that tend to dominate are neither voting nor money,
but whether a bug is easy to fix.  If you have a bug, you can make it more
likely that it will get fixed by helping to make it easy to fix, by producing
a really good bug report.  A good bug report is one that provides complete
information and a short testcase that reliably triggers the bug.

That's why folks like Wolfgang Bangerth, who specialize in producing reduced
test cases, are GCC heros.

Re: bugzilla voting

[Phil Edwards]

On Fri, Apr 02, 2004 at 11:03:40AM -0800, Joe Buck wrote:
> 
> That's why folks like Wolfgang Bangerth, who specialize in producing reduced
> test cases, are GCC heros.

Hear, hear!

-- 
Behind everything some further thing is found, forever; thus the tree behind
the bird, stone beneath soil, the sun behind Urth.  Behind our efforts, let
there be found our efforts.
              - Ascian saying, as related by Loyal to the Group of Seventeen

Re: bugzilla voting

[Robert Dewar]

Scott Robert Ladd wrote:

> I would like to see a voting system, so long as there are reasonable 
> assurances that the vote has some value. I don't expect the tyranny of 
> the commons to drive every decision by the SC or release manager -- I 
> would, however, want to see respect for the desires of the user community.

I doubt that the voting is an accurate indication of user community 
desires, and in fact I see no reason to pay attention to it. People who 
would= rather spend energy on voting on what others should volunteer to 
do, rather than volunteering themselves, are not people to whose 
opinions one would expect a high degree of attention to be paid :-)

So I think you are lacking the assurance you require!

Re: bugzilla voting

[Robert Dewar]

Per Abrahamsen wrote:

> Scott Robert Ladd <coyote@coyotegulch.com> writes:

> That sentiment would be an argument against enabling voting.  I
> certainly would not want to give assurances any stronger than that I
> just *might* take votes into account occasionally.

If we have voting then the only thing the vote indicates is
that people have entered a certain number of votes. We cannot
legislate, urge, predict, or depend on anyone reacting in
any particular way to this information. It may help increase
the priority of a bug, it may have the opposite effect, who
knows?

Re: bugzilla voting

[Robert Dewar]

Scott Robert Ladd wrote:

> Please note that I am not criticizing Pers! A purely democratic approach 
> won't work, either, as under such, the only bugs that would get fixed 
> are those for very popular platforms.

Also, volunteers are not elected officials, They are not working for
users, and cannot be ordered around :-) The most you can do is
encourage them, and perhaps voting will provide that encouragement
for some developers.