bug reporting

[Richard Frith-Macdonald]

On 13 Feb 2007, at 23:59, Chris Vetter wrote:

> On 2007-02-13 21:40:26 +0100 Richard Frith-Macdonald  
> <richard@tiptree.demon.co.uk> wrote:
> > I suspect that comments like that do a lot to put developers off   
> > working on GNUstep, doing immeasurably more harm than breakage of   
> > code in svn-trunk.
> > The implied notion that a change must be tested on all platforms   
> > before submission to svn-trunk in order to qualify as  
> > demonstrating  'some care' is frankly ridiculous and insulting.
>
> No. I'm sorry, but no.
>
> I'm NOT saying that code needs to be tested on every possible  
> platform. That would indeed be ridiculous.
>
> I'm saying that you should use a bit of brain-work before commiting  
> code. Using a hard-coded call to make(1) simply ASSUMING that said  
> make(1) IS actually 'GNU make' is, pardon my French, stupid.

Well, the issue is not really whether you are saying that people  
should have tested thoroughly on all platforms or should have spotted  
all errors by using 'a bit of brain work'  is it?  It's whether it's  
appropriate and productive to react to bugs in software contributions  
put on svn-trunk by insulting/swearing.

The whole point of svn-trunk is to provide a repository where a large  
number of people can test out work,  so that we end up with reliable  
and portable code before a release rather than fixing lots of bugs  
after a release.  It is unreasonable to expect code submitted for  
testing to be perfect ... indeed it's unreasonable not to expect it  
to contain major and glaringly obvious errors on occasion... most  
errors are obvious once spotted, and everyone makes stupid mistakes  
sometimes.

For svn to work effectively in that role, It is essential that  
updates should be made in relatively small, frequent batches ... so  
that the revision control can be used to roll-back to earlier  
versions and localise problems.  The use of frequent comittal means  
that instead of one core developer spending a huge amount of time  
testing and considering  a load of changes, lots of less experienced  
people can spot problems and localise them so they can be fixed.

The responsibility of people using svn-trunk is to test, spot errors,  
and politely report them with plenty of detail/explanation  
(preferably with a fix) or just fix them.
The incentive is that they get to play with new features early, and  
they get to know that they are helping the project progress.
And using svn, you never have to wait for someone to provide a fix  
after submitting your bug report... if you can't (lack of skill or  
time or inclination) fix a problem yourself, you just use the  
previous revision.