tech-pkg archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: [SORRY] I broke python311 and other packages



Aleksey Cheusov <vle%gmx.net@localhost> writes:

> Hi everyone. Recently I broke some packages trying
> to add/improve "nls" and "readline" options to packages.
>
> First of all I am very sorry for this :-/
>
> Can any one please explain what's wrong with my changes.  It seems
> that I'm missing something fundanetal but I don't see where exactly.

I am unclear on the details, so some meta comments.

  For every commit, there should be an expectation that every package that
  worked before will continue to work, on every platform where it
  previously did, including for platforms that you didn't/can't test on.
  Of course occasionally it doesn't come out this way, but it should
  mostly.

  The above implies that you should at least be building and testing on
  some platform.

  If you can't predict that there will be esesntially no objections, the
  change should be proposed/discuss first.  python is a bit odd in that
  it has MAINTAINER=pkgsrc-users but it's obviously hugely important.

  Adding an option to do something that the package didn't do before,
  and that defaults off, is not a big eal.

  Adding an option that splits something that used to be present into an
  option, and defaults on, is not a big deal.  It's essentially "add an
  option to turn it off".

  Adding something and defaulting on is almost equivalent to just adding
  it, and that may or may not need discussion.  For python, it probably
  does.

  In my view (and this may be stricter than community standards), it
  should be clear from the commit message what kind of options change is
  happening.  The reason should be given too.  e.g. "Readline was always
  enabled before, and now people can disable the option to avoid the
  code size or the license.  There should be change in the default
  build".  That allows people to review/test the change against the
  intent.

  When commiting, one needs to read email sent to user%netbsd.org@localhost
  reasonably promptly.  I saw objections followed by more commits.


Home | Main Index | Thread Index | Old Index