Update on bug system for OpenJDK (web-discuss)
Roger Lewis
roger.lewis at oracle.com
Thu Jun 2 19:56:28 UTC 2011
>> I agree - but that needs to be handled by the "first tier" system (as
>> is mostly done today). I don't want to be the one who has to wade
>> through these issues and identify them - by the time a bug gets into
>> bugster today it "should" have reasonable merit and shouldn't be a
>> duplicate.
> It's not clear to me that we need this filtering, at least not
> initially. I can't imagine end-users seeking out the OpenJDK bug
> database, unless they have a message dialog or error log that tells
> them to go there. So if consumers could be directed somewhere else
> then there's a good chance that most of the bugs will come from
> developers that want to help by reporting a bug or maybe contributing
> a fix. Sure, we'll still get some poor quality bug reports but that is
> par for the course. I would suggest it would be better to see how it
> goes before deciding to setup a quarantine area. Another thing is that
> one would hope that an open bug database will encourage more
> volunteers to help, and so there should be more eyes on these bug
> reports (and so one would hope that the poor reports will be closed
> quickly).
Filtering at this level, to some degree, can be achieved by providing an
appropriate reporting path to users being targeted. Not all submission
paths suit all users.
For Java SE as it is today, all bug reporting occurs on
bugreport.sun.com, but the depending in the context you see this domain,
the page you are directed to is intended to obtain the information that
is necessary to make a useful report, and present the reporters with
questions that they are likely able to answer. Also, the resulting
location of those reports can vary depending on the path.
- For end users, on java.com, the page they are directed to:
http://bugreport.sun.com/bugreport/main.jsp
- Java Developers/IT administrators/'more advanced end users', and
others, who find this next link in ReadMe's, on bugs.sun.com,
documentation, and elsewhere are provided with :
http://bugreport.sun.com/bugreport/
Within that reporting path, the questions can vary for bug vs RFEs
and in some cases vary on the 'product' the reporter selects.
This path 'loosely' requires fields like 'java -version', a test
case, steps to reproduce, what you expected to happen and what
really happens. This results in a report that looks like this, where
all of the output of those fields are 'stuffed' into a description
field in BugTraq:
http://bugs.sun.com/view_bug.do?bug_id=6550655
- The most commonly used reporting URL (around 450 reports a month)
is presented to users when they experience a HotSpot crash:
http://java.sun.com/webapps/bugreport/crash.jsp which redirects to
http://bugreport.sun.com/bugreport/crash.jsp
In this reporting path there is an effort made to separate out end
users who have experienced a crash vs. those who are developers who
have the ability to diagnose and report a full and complete 'bug
report'.
- For those who can access BugTraq, they are able to create bug
reports that do no go into a 'triage' area and are not forced to go
through the lengthy set of questions.
This set of paths is not perfect by any means. Overtime, with adding and
improving paths, we have seen bug quality rise significantly over the
single text box that appeared on what was the original reporting page:
http://java.sun.com/cgi-bin/bugreport.cgi
Given the vast user base of Java being able to provide the right users
with an appropriate and customizable path is essential.
Anonymous reporting into a non-triage area, should not be allowed, as
those paths should provide a less structured reporting process. In some
cases, for some, like OpenJDK contributors, there should be only a
minimum set of data necessary to create a report.
>
>> :
>>
>> I was with you right up to the last sentence. Having to register
>> would dissuade some users from bothering to report non-bugs (but it
>> may also dissuade them from submitting real bugs!). There needs to be
>> a system where a report can easily be entered by anyone, but that
>> report needs to be pre-processed before being accepted as a bug (in
>> todays terms it gets submitted as an "incident", processed and if
>> deemed suitable then a bug is created in bugster).
> I don't know how many projects allow bug reports to be submitted
> anonymously. At least with bugs.openjdk.java.net you have to login,
> and same thing for other bugzillas that I've submitted bugs to. With
> the legacy bugs.sun.com then you also have to fill out a form with
> your details. One useful thing about requiring a submitter to login is
> that there is contact address, useful if follow-up is required to
> duplicate or verify a bug.
Bug reporting on bugs.sun.com is somewhat anonymous...... Which has lead
to many hallway discussions as to whether it should be allowed. In the
end is what is there today. An email address is mandatory, that email
address is only validated in that is matches a
name at domain.toplevelDomain format. The thought has been that people do
not want to create a new account in a system to report a bug, and we
want to make the process easy enough to not scare off those who do not
want to go through that step. Though having a login, in some submission
paths, would be nice, it was unfortunately never implemented on
bugs.sun.com.
Most reporters provide valid addresses, though we do see fake email
addresses, that bounce. Often, even if they have a corporate address,
say name at oracle.com, they will use their personal, or a
name-spam at domain.com address.
-Roger
More information about the discuss
mailing list