[repo-coord] Packaging Process - PLEASE READ

[Bent Terp]

> not every source download provides a gpg signed or md5 checked source.
> we need a general solution for that.

We cannot check the integrity of a tarball unless it has some sort of
checksum. That work has to be done upstream.

>> > 2. before you even touch the source check if its already packaged.
>>
>> Bien sûr, life is short.
>
> so we rather break things and override ... if the work is already done
> you safe time...

Maybe I wasn't clear here, but I agree 110% with you. I meant that my life
is too short to be writing spec files if somebody else has already done it
for me.

> send it upstream
> and it can be fixed... everything else is butchering with alot bork
> potential. especially if just blindly done on a new release.

Usually, I don't do bugfixes - that's the authors job. I mainly use perl
one-liners to deal with authors' peculiar ways of writing makefiles. You
may not have come across situations like I have, but I can assure you that
these scientific types have some strange ideas about make should be done
during 'make install' - this is far removed from ./configure
standardisation. And sending makefile.diff upstream isn't going to make
these professors change their minds....


> and that is really a problem for you? if you just look in their macro
> definitions its no problem at all ;)...

I haven't been building that long, and find redhat's macros confusing
enough. I'm a bit reluctant to risk what little sanity I have left on
trying to understand other distro's specific macros.

>> > 1. building should always happen in a non root environment
>>
>> This sentence uses the word "should" with the same strength as in an
>> RFC: "should" = "absolutely always must". Which is why I haven't rebuild
>> OpenPBS from the oscar.sf.net distribution, yet.
>
> read the docs on rpm.org

Again, I never meant to sound as if I disagree with you, cuz I don't. And
so I havent released OpenPBS yet, simply because I can't get the darned
thing to build as non-root.

> if we dont check we can just use the borked CPAN stuff... why package it
> all then?

Because my users demand i.e. the latest bioperl, and they don't care if
'make tests' gives some errors as long as the modules they need, work.

> lots of funny comments yet... anything productive?

Not unless you consider it productive to hear about my
problems/experiences. I gather you dont, so I apologoise for wasting even
more of your time.

-- 
/Benty