www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - bugzilla template

reply "Steven Schveighoffer" <schveiguy yahoo.com> writes:
Many times, a new user of D (and even some experienced ones) will post a  
bug to bugzilla with not enough or just enough information, but could be  
way more informative.

For example, take http://d.puremagic.com/issues/show_bug.cgi?id=6323

I have to download this code and compile it, to see what he is talking  
about (no error messages or explanation of what's wrong).  (sorry to pick  
on this bug, it's not the only one, but the most recent)

I think requiring someone to download code and compile it is way too much  
overhead to determine whether a bug is trivial, duplicate, invalid, etc.

In other bug reporting systems I've used, when you create a new bug, it  
gives you a template to fill out (i.e. steps to reproduce, any relevant  
error messages).  I think this would go a long way to getting more  
informative bug reports.

Is it possible to add something like this to D's bugzilla?  Do other  
people think this is a good idea?

I'll give a sample:

-------------------------------------
Please answer the following mandatory questions

1. List an example that causes the bug (please reduce as much as  
possible).  Do not paste more than 100 lines of code.

2. If relevant, list the command line used to compile, including any flags  
passed to the compiler

3. If any errors are output by the compiler, list them here

4. What is your expected result?

5. What is the actual result?

Note: please use the selectors above to fill out as much detailed  
information as possible (OS, architecture, version of D, etc.)
-------------------------------------

I'm not stuck on this form, probably the most prolific bug fixers could  
come up with a better one.

-Steve
Jul 15 2011
next sibling parent reply "Vladimir Panteleev" <vladimir thecybershadow.net> writes:
On Fri, 15 Jul 2011 18:45:15 +0300, Steven Schveighoffer  
<schveiguy yahoo.com> wrote:


 4. What is your expected result?

 5. What is the actual result?


I hate bug reporting forms that FORCE you to answer these questions. Most  
of the time they will be along the lines of "4. It should work" and "5.  
The error message I specified above appears". A template is nice, but I  
would avoid saying that answering all questions is mandatory unless simply  
listing them as a guideline shows to be ineffective.

-- 
Best regards,
  Vladimir                            mailto:vladimir thecybershadow.net
Jul 15 2011
parent reply "Steven Schveighoffer" <schveiguy yahoo.com> writes:
On Fri, 15 Jul 2011 13:34:53 -0400, Vladimir Panteleev  
<vladimir thecybershadow.net> wrote:


 On Fri, 15 Jul 2011 18:45:15 +0300, Steven Schveighoffer  
 <schveiguy yahoo.com> wrote:


 4. What is your expected result?

 5. What is the actual result?


 I hate bug reporting forms that FORCE you to answer these questions.  
 Most of the time they will be along the lines of "4. It should work" and  
 "5. The error message I specified above appears". A template is nice,  
 but I would avoid saying that answering all questions is mandatory  
 unless simply listing them as a guideline shows to be ineffective.


It would not be enforced, it would be the text that appears by default in  
the description.  Of course, you can just delete all that and write your  
own, or you could just ignore the questions I suppose.

It's just that we get bug reports like:

Summary: dmd doesn't compile this

Description:

void main()
{
    writefln("hello world");
}


With nothing else.  Who freaking knows why this doesn't compile on X's  
system?  But with an error message like:

Error: writefln not defined

Then I don't have to guess, it's close as invalid, takes about 10 seconds  
out of my life instead of 10 minutes.

The expected/actual is more for cases where the code is doing something  
you don't expect like "I expected this to print 0 but it prints 1  
instead"  Again, bug reports come in where the code is the only clue, and  
if the code works as defined, it's hard to determine what is perceived as  
wrong.

I'm thinking something like this would reduce the amount of time it takes  
to debug the trivial bug reports (invalid, works as defined, duplicate,  
etc.).

-Steve
Jul 15 2011
parent reply "Vladimir Panteleev" <vladimir thecybershadow.net> writes:
On Fri, 15 Jul 2011 21:34:59 +0300, Steven Schveighoffer  
<schveiguy yahoo.com> wrote:


 It would not be enforced, it would be the text that appears by default  
 in the description.  Of course, you can just delete all that and write  
 your own, or you could just ignore the questions I suppose.


Yes, I understand. My point was mainly about using the word "mandatory" in  
your template suggestion.

-- 
Best regards,
  Vladimir                            mailto:vladimir thecybershadow.net
Jul 15 2011
parent "Steven Schveighoffer" <schveiguy yahoo.com> writes:
On Fri, 15 Jul 2011 15:01:03 -0400, Vladimir Panteleev  
<vladimir thecybershadow.net> wrote:


 On Fri, 15 Jul 2011 21:34:59 +0300, Steven Schveighoffer  
 <schveiguy yahoo.com> wrote:


 It would not be enforced, it would be the text that appears by default  
 in the description.  Of course, you can just delete all that and write  
 your own, or you could just ignore the questions I suppose.


 Yes, I understand. My point was mainly about using the word "mandatory"  
 in your template suggestion.


Yes, you are right, It should probably not say mandatory.

-Steve
Jul 15 2011
prev sibling parent reply KennyTM~ <kennytm gmail.com> writes:
On Jul 15, 11 23:45, Steven Schveighoffer wrote:

 Many times, a new user of D (and even some experienced ones) will post a
 bug to bugzilla with not enough or just enough information, but could be
 way more informative.

 For example, take http://d.puremagic.com/issues/show_bug.cgi?id=6323

 I have to download this code and compile it, to see what he is talking
 about (no error messages or explanation of what's wrong). (sorry to pick
 on this bug, it's not the only one, but the most recent)

 I think requiring someone to download code and compile it is way too
 much overhead to determine whether a bug is trivial, duplicate, invalid,
 etc.

 In other bug reporting systems I've used, when you create a new bug, it
 gives you a template to fill out (i.e. steps to reproduce, any relevant
 error messages). I think this would go a long way to getting more
 informative bug reports.

 Is it possible to add something like this to D's bugzilla? Do other
 people think this is a good idea?

 I'll give a sample:

 -------------------------------------
 Please answer the following mandatory questions

 1. List an example that causes the bug (please reduce as much as
 possible). Do not paste more than 100 lines of code.

 2. If relevant, list the command line used to compile, including any
 flags passed to the compiler

 3. If any errors are output by the compiler, list them here

 4. What is your expected result?

 5. What is the actual result?

 Note: please use the selectors above to fill out as much detailed
 information as possible (OS, architecture, version of D, etc.)
 -------------------------------------

 I'm not stuck on this form, probably the most prolific bug fixers could
 come up with a better one.

 -Steve


It's getting worse: http://d.puremagic.com/issues/show_bug.cgi?id=6327.

Anyway, item 5 shouldn't be mandatory. The actual result is in item 3. 
And even item 4 is optional. The expected result is item 3 shouldn't happen.

There is actually this page:

    http://d.puremagic.com/issues/page.cgi?id=bug-writing.html

I believe no one had read this :)
Jul 15 2011
parent reply "Steven Schveighoffer" <schveiguy yahoo.com> writes:
On Fri, 15 Jul 2011 16:25:25 -0400, KennyTM~ <kennytm gmail.com> wrote:


 On Jul 15, 11 23:45, Steven Schveighoffer wrote:

 3. If any errors are output by the compiler, list them here

 4. What is your expected result?

 5. What is the actual result?

 Note: please use the selectors above to fill out as much detailed
 information as possible (OS, architecture, version of D, etc.)
 -------------------------------------

 I'm not stuck on this form, probably the most prolific bug fixers could
 come up with a better one.

 -Steve


 It's getting worse: http://d.puremagic.com/issues/show_bug.cgi?id=6327.

 Anyway, item 5 shouldn't be mandatory. The actual result is in item 3.  
 And even item 4 is optional. The expected result is item 3 shouldn't  
 happen.


There are cases where the code does compile, and outputs the wrong  
result.  They could have nothing to do with the compiler.



     http://d.puremagic.com/issues/page.cgi?id=bug-writing.html

 I believe no one had read this :)


Nobody probably knows it exists (including myself until now!)

-Steve
Jul 15 2011
parent KennyTM~ <kennytm gmail.com> writes:
On Jul 16, 11 05:02, Steven Schveighoffer wrote:

 On Fri, 15 Jul 2011 16:25:25 -0400, KennyTM~ <kennytm gmail.com> wrote:


 On Jul 15, 11 23:45, Steven Schveighoffer wrote:

 3. If any errors are output by the compiler, list them here

 4. What is your expected result?

 5. What is the actual result?

 Note: please use the selectors above to fill out as much detailed
 information as possible (OS, architecture, version of D, etc.)
 -------------------------------------

 I'm not stuck on this form, probably the most prolific bug fixers could
 come up with a better one.

 -Steve


 It's getting worse: http://d.puremagic.com/issues/show_bug.cgi?id=6327.

 Anyway, item 5 shouldn't be mandatory. The actual result is in item 3.
 And even item 4 is optional. The expected result is item 3 shouldn't
 happen.


 There are cases where the code does compile, and outputs the wrong
 result. They could have nothing to do with the compiler.


Yes, you can always include additional information. And change item 3 to 
"errors output by the compiler or the executed program".


 http://d.puremagic.com/issues/page.cgi?id=bug-writing.html

 I believe no one had read this :)


 Nobody probably knows it exists (including myself until now!)

 -Steve
Jul 15 2011