How much time will D1 be around?

[Brad Roberts]

Jarrett Billingsley wrote:
> On Tue, Nov 11, 2008 at 10:44 PM, Brad Roberts <braddr at puremagic.com> wrote:
>> Jarrett Billingsley wrote:
>>> It's not like the DMDFE can get any _worse_.
>> I try to avoid responding to you for the most part, but when you make
> 
> Wow, I wasn't aware you had this kind of disdain for me.  I'd be
> interested to know why.  Yeah, I know, as you're reading this you're
> palming your forehead and going "wow! how can he be so oblivious!"
> But no, seriously.

I overstated it, but mostly to drive home the point.  It's primarily the
level of negativity you show on most threads that mention compiler bugs.
 You just can't seem to help but exaggerate and use antagonistic phrasing.

>> statements that are so obviously uncalled for and egregiously rude, I
>> can't help it.
>>
>> This is over the top out of line, imho.
> 
> My comment was not really meant to provoke, but I suppose that it
> could be taken that way, and for that I apologize.

How in the world can a statement like that NOT be taken that way?  DMD
is obviously a very usable compiler.  It could get a hell of a lot worse.

> The point I'd've liked to have made is that yes - Walter has
> (obviously!) done a lot for D.  Without him, D would not exist.  But
> the fact is - and I doubt you'll find many who will disagree with me -
> it doesn't matter how awesome Walter is, or how much of a visionary he
> is, or how good his managerial skills are; DMDFE is just buggy, no
> offense meant to Walter as a person.  That there are 30 new bugs
> posted to the bugzilla every month from a relatively small group of
> users is testament that DMDFE is buggy.  And the current development
> model of "have people put things in bugzilla where there is a very
> good chance that they will never get fixed" and "be extremely
> skeptical of any and all patches that people submit" does not work.
> As I said - simple flow problem.  30 in, 15 out.  Eventually the tank
> is going to overflow.

Depends on your definition of buggy.  If it's the extreme of "there is
at least one known / unfixed bug" then yes, it's buggy.  But so is every
other compiler produced for every other language out there.  Hardly a
useful definition.

As to "have people put things in bugzilla where there is a very good
chance that they will never get fixed".. what do you suggest people do?
 Keep them to themselves and hope they get fixed?  Hope someone else
stumbles into the same bug and files it for them?  Those are certainly
good ways of dealing with them (dripping sarcasm, yes).

I won't try to speak for why bugs with patches haven't all been
resolved, since I have no extra insight into Walter's brain.  I to wish
they'd get more attention due to the symbiotic relationship they build.

So.. seemed like it might be interesting to pull some stats from
bugzilla.  For all of the below, the data is restricted to bugs filed
for the product 'D' (leaving out dgcc, dmc, dstress, puremagic.com, and
testproduct).  Filed means any status.  Resolved means status is
resolved, verified, or closed.  Open means status is unconfirmed, new,
assigned, or reopened.

An important point before getting into the actual numbers.  Raw numbers
are just that.. raw.  Not all bugs are of equal weight.  The fact that
there are open bugs is NOT a sign of anything other than that there open
issues.  It doesn't provide any way to judge the usability or quality of
the code upon which the bugs have been reported.  There's _some_
qualitative measure based on severity, but as this field isn't terribly
actively/accurately maintained in d's bugzilla, it's risky to use it
though I did break out the stats for that a little in the numbers below.

Per month for all versions:
  jan 2008: 54 filed, of which 30 are resolved
  feb 2008: 65 filed, of which 37 are resolved
  mar 2008: 70 filed, of which 15 are resolved
  apr 2008: 87 filed, of which 28 are resolved
  may 2008: 68 filed, of which 33 are resolved
  jun 2008: 47 filed, of which 18 are resolved
  jul 2008: 67 filed, of which 25 are resolved
  aug 2008: 62 filed, of which 28 are resolved
  sep 2008: 55 filed, of which 22 are resolved
  oct 2008: 52 filed, of which  8 are resolved
  nov 2008: 12 filed, of which  1  is resolved


Some 0.x and 1.x combined stats:
  390 open prior to 2008
  908 resolved prior to 2008

  137 open for 2008
   99 resolved for 2008
-----
527 of 1534 bugs still open.  That's just over 34% open vs 66% resolved.
That's a far cry from "there is a very good chance that they will never
get fixed".  Even if you look at just 2008's percentages, 58% open vs
42% resolved.  Not as good, but still not super bleak.


Some 2.x stats:
  75 opened 2.x bugs prior to 2008
  75 resolved 2.x bugs prior to 2008

  215 opened 2.x bugs during 2008
  111 resolved 2.x bugs during 2008
-----
290 of 476 bugs still open.  That's just shy of 61% open vs 39% resolved.


Taking a look at the stats for the bugs with severity regression,
blocker, or critical:
  30 (24%) open 0.x/1.x
  94 (76%) resolve 0.x/1.x

  15 (33%) open 2.x
  30 (67%) resolved 2.x


Bugs with patches attached and marked as patches:
  10 (26%) open 0.x/1.x
  29 (74%) resolved 0.x/1.x

   8 (67%) open 2.x
   4 (33%) resolved 2.x

There's 239 filed issues that don't specify a version.  Of those, 118
(49%) are open and 121 (51%) resolved.  So not wildly different from the
above data.

Well, enough searching of bugzilla for tonight.

Later,
Brad