• Graham Fawcett <fawcett uwindsor.ca> Jun 15 2010
• bearophile <bearophileHUGS lycos.com> Jun 15 2010
• Don <nospam nospam.com> Jun 15 2010
• Graham Fawcett <fawcett uwindsor.ca> Jun 15 2010
• Tomek =?UTF-8?B?U293acWEc2tp?= <just ask.me> Jun 15 2010
• BCS <none anon.com> Jun 15 2010
• BCS <none anon.com> Jun 15 2010

Graham Fawcett <fawcett uwindsor.ca> writes:

Hi folks,

Would you be interested in starting a "bug-of-the-week club" on the
discussion list?

This list seems to have a good culture of submitting bug reports, and
of referring the bug-bitten to relevant existing reports. But many of
the reports seem to sit on the buglist for several months (or years).
I suspect the core team is too small to adequately address them all in
a timely manner. A semi-formal community process might help.

I propose that the BOTW club (lifetime memberships for all!) selects
an arbitrary bug once every week, and delves into it collectively.
This might lead to positive results: shorter times-to-fix; more
frequent bugfix releases; the contribution of unit tests to prevent
regressions; and a broader understanding of D internals, which might
foster the growth of the core team.

Thoughts?
Graham

bearophile <bearophileHUGS lycos.com> writes:

Graham Fawcett:
 Would you be interested in starting a "bug-of-the-week club" on the
 discussion list?


A focus on a "hard bug of the week" can be a good thing to hopefully remove
some important old bugs.

Bye,
bearophile

Don <nospam nospam.com> writes:

Graham Fawcett wrote:
 Hi folks,
 
 Would you be interested in starting a "bug-of-the-week club" on the
 discussion list?


In a sense, something like that already exists. Anyone who is interested 
in helping compiler bugs get fixed should join the dmd-internals mailing 
list.

BTW - It's worth knowing that a significant part of the effort in fixing 
a bug (maybe 25% of the total time) is in creating a minimal test case. 
I've noticed that very few bug reports have completely minimal test 
cases. Reducing existing test cases is something that anyone can do.

Graham Fawcett <fawcett uwindsor.ca> writes:

On Tue, 15 Jun 2010 21:56:57 +0200, Don wrote:

 Graham Fawcett wrote:
 Hi folks,
 
 Would you be interested in starting a "bug-of-the-week club" on the
 discussion list?


 In a sense, something like that already exists. Anyone who is interested
 in helping compiler bugs get fixed should join the dmd-internals mailing
 list.


Thanks, Don! I'll take a look at the dmd-internals list.

 BTW - It's worth knowing that a significant part of the effort in fixing
 a bug (maybe 25% of the total time) is in creating a minimal test case.
 I've noticed that very few bug reports have completely minimal test
 cases. Reducing existing test cases is something that anyone can do.


+1.

best,
Graham

Tomek =?UTF-8?B?U293acWEc2tp?= <just ask.me> writes:

Don wrote:

 BTW - It's worth knowing that a significant part of the effort in fixing
 a bug (maybe 25% of the total time) is in creating a minimal test case.
 I've noticed that very few bug reports have completely minimal test
 cases. Reducing existing test cases is something that anyone can do.


Maybe bug writing guidelines specifically tailor-made for compiler test 
cases would be a good idea (e.g. "Make sure if someone else copy-pastes your 
code, it will compile (or fail in the way described).", etc). I find the 
bugzilla guidelines a bit too generic. Another thing is that they could be 
more "in your face" (pre-filled description?); the link at the top is hard 
to notice.

-- 
Tomek

BCS <none anon.com> writes:

Hello Don,

 BTW - It's worth knowing that a significant part of the effort in
 fixing a bug (maybe 25% of the total time) is in creating a minimal
 test case. I've noticed that very few bug reports have completely
 minimal test cases. Reducing existing test cases is something that
 anyone can do.


Maybe the practice of including a minimal test case as an attachment on bugs 
should be adopted. If it was done in a way that can be detected automatically, 
them bugs without them could be filtered for and the test cases scraped into 
a test suit (so we can tell when a bug vanishes on it's own). For that matter 
what would it take to run all .d attachments on open bugs thought each new 
version of DMD and attache the results?
 
-- 
... <IXOYE><

BCS <none anon.com> writes:

Hello Graham,


 I propose that the BOTW club (lifetime memberships for all!) selects
 an arbitrary bug once every week, and delves into it collectively.
 This might lead to positive results: shorter times-to-fix; more
 frequent bugfix releases; the contribution of unit tests to prevent
 regressions; and a broader understanding of D internals, which might
 foster the growth of the core team.
 
 Thoughts?


If Walter would agree to give detailed public feedback on BOTW patches (It's 
only one patch a week) it might improve the quality of all patches.

-- 
... <IXOYE><