[ANNOUNCE] dh_splitpackage 0.1

Zygmunt Krynicki zygmunt.krynicki at linaro.org
Fri Jun 3 22:55:23 UTC 2011

W dniu 03.06.2011 21:17, Evan Broder pisze:

> Ubuntu has gotten quite a bit of flack from the debhelper maintainer
> for making independent changes in the past ([1], [2]), and doing so
> again seems like a bad plan.

Perhaps posting this from my @canonical.com address was a bad idea. I'm 
not doing this as a Canonical employee, nor as a ubuntu developer. I'm 
doing this as an upstream (actually working on Linaro tools) that uses 
Debian/Ubuntu as the delivery platform.

My response below may seem to be quite harsh but I hope to keep it factual.

For context. I wrote dh_splitpackage yesterday after being upset with 
dh_install and failing to find any packages that would do what I wanted 
easily. I did it while packaging the tools I'm developing daily so that 
my users can install them easily. Nothing in this has any relation to my 
employer, Canonical, Ubuntu or Linaro. I made this and the fault, if 
any, is mine.

> dh_install already has a --fail-missing option. It seems that this
> would cover half of your use case,

To the best of my knowledge dh_install cannot do what I wish. Well, I 
could copy file by file, directory by directory, carefully observing the 
right paths, adding exclusion patterns and so on.

That's not my goal.

My goal is to split a package (which I understand as "move each file 
from debian/tmp into the appropriate sub-package") easily, without too 
much, error-prone, shell scripting and without doubts I missed something 
and it's actually failing silently (or will fail silently when upstream 
files move around later).

> though I gather there are some
> outstanding issues with it ([3]). CDBS also has a list-missing target,
> though it doesn't seem that it can be made to fail the build, and also
> doesn't cover your overlapping files issue.

See, that's not solving my problem.

> It seems like it would be much more productive to work with Joey to
> try and fix dh_install --fail-missing, and possibly add a new
> --fail-on-overlap option or something.

Productive for who? I never even knew Joey (no disrespect to him). I 
solved my problem and shared the solution. My problem is solved. Better 
yet, I did not make the general problem worse.

Look at this as a contribution to possible pool of solutions that the 
greater Debian community may choose to adopt.

> But in any case, I think that
> adding a new debhelper script without consulting Joey at all will hurt
> Ubuntu's image in Debian's eyes, and that's a bad thing.

I'm sorry I did not know Joey. Perhaps if I was working on packaging 
lots of 3rd party software I would get know him eventually. Should that 
prevent me (or anyone) from coming up with a good technical solution to 
a problem? Even if dh_splitpackage will rot in the internet archive it 
_did_ solve my problem and I _did_ share the solution hoping for the best.

Now for some rants of my own:

Let's play a game that I did knew Joey or other dh_* maintainers.
IMHO engaging in platform development would be pointless. It might end 
up being political. It would most likely drag on and on, possibly never 
reaching a consensus. It might require me to rewrite the script in perl, 
shell or another language. And it could easily fail.

Why would  I engage in any of that? Why would any upstream do? (and see 
how friendly upstream I am to even share my solution with anyone, do you 
think random upstream developer would?). Debian is an insanely complex 
political being, with many rules that are fully versed only by people 
that dwell in this environment for many years. That's not a good ISV 
environment. Perhaps this is not a popular view but IMHO it's rather 
true (why would I be writing this email otherwise?)

Getting involved in fixing the platform the way Debian maintainers and 
many of my colleagues at Canonical do would be diverging from my goals. 
I'm an upstream working on a project. This project is not Debian. 
Getting involved with Debian, dh_* maintainers, authors and 
contributors, while noble, would prevent me from attaining my goal on a 
predictable and timely basis, which is to get the software I'm working 
on into the hands of my users.

End of rant.

Now I understand that you had no bad intentions towards me and this is 
all caused by me posting from @canonical.com without giving any kind of 
explanation on why I did this.

Zygmunt Krynicki

More information about the ubuntu-devel mailing list