[Python-Dev] User's complaints

[Guido van Rossum]

On 7/12/06, Brett Cannon <brett at python.org> wrote:
> On 7/12/06, Guido van Rossum <guido at python.org> wrote:
>
> > On 7/12/06, Armin Rigo <arigo at tunes.org> wrote:
> > > I guess I'm going to side with Greg Black on his blog entry.
> >
> > I seem to recall that that particular one wass *not* an accidental
> > bug. I believe I fell over exactly the problem that Greg Black
> > complained about (or almost the same; maybe my problem was setting the
> > month and day to zero and formatting only the time :-) and tried to
> > convince the author to change it; but was told that the new behavior
> > was never documented and that the new behavior was somehow better; and
> > I didn't fight it further then. Do I remember this correctly? Does
> > anybody else recall?
>
>
> My recollection is that we found a way to cause a crash if improper values
> were used.  We never said 0s were allowed in the docs and that it could mask
> bugs if you did use them and we supported them ( e.g., setting 0 for January
> instead of 1 if you were thinking in terms of indexing at the time).  So we
> said that values should be within the proper range and not above or below
> them.
>
> The python-dev Summary coverage can be found at
> http://www.python.org/dev/summary/2004-02-01_2004-02-29/#time-strftime-now-checks-its-argument-s-bounds

Thanks for confirming  memory! So it was an intentional regression;
"bugs happen" doesn't apply in this case. And an unfortunate
regression at that -- not because one guy writes a silly blog entry
about it, but because it breaks real code -- undocumented be damned.

Are we going to fix it, without allowing those crashes again?

-- 
--Guido van Rossum (home page: http://www.python.org/~guido/)