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

Re: dpkg 2.0?

Esteban Manchado Velázquez writes ("Re: dpkg 2.0?"):
> On Wed, Oct 04, 2006 at 12:03:39PM +0100, Ian Jackson wrote:
> > The dpkg code is quite old, and badly abused by one particularly
> > prolific previous maintainer, so it is in need of an overhaul by an
> > expert C programmer.
>     Sorry for insisting so much, but I really think we should try to get at
> least some basic dpkg testing _before_ we refactor code or make any
> "overhaul".

I think I have generally less of a high opinion of testing in areas
where we have high levels of expertise.

My experience as upstream maintainer for adns (which has a pretty
comprehensive regression test suite) doesn't encourage me to think
highly of tests: the fact that adns testing is done with a regression
suite has not ever prevented me from shipping a bug.  Of course ad-hoc
testing has caught bugs, and usually I would add the ad-hoc test for a
change to the test suite, but it's not clear what the value of the
tests _as regression tests_ is.

>     I don't know about you, but I'm _very afraid_ of touching badly
> written C code without any "safety net". Especially if that code
> belongs to the most important program in Debian.

I think I can safely say that, with the code for dpkg in front to me,
I am generally able to determine both the intended and actual
behaviours, and able to produce correct changes.

But perhaps you have misunderstood what I mean by `overhaul'.  I don't
propose to generate a flag day with a hugely different set of code.
Rather, the way to do the overhaul would be as a series of changes
each of which fixes a particular set of problems; after each change is
made, we can stop and see how we're doing.


Reply to: