Am 02.04.18 um 12:41 schrieb Johnny A. Solbu:
> On Tuesday 13. March 2018 16.37, Christoph Lechleitner wrote:
>> Yep, a simple
>> dpkg-buildpackage -b -us -uc
>> ran through without issues.
>>
>> I'll probably test the resulting package on our own server soon,
>> with nighly backups as insurance.
>
> first, sorry for not seeing this earlier.
Don't be, it wouldn't have made any difference to my choices ;-)
> If there is going to be an upstream repo for everyone to use,
Our's is an upstream-based repo, not an upstream repo, and I think that's clear enough.
Keeping our repo private would be wrong.
> I highly advice against building binary only packages.
With arch independent there's no real difference between binary and source packages ;-)
> (The «-b» switch)
In our case there'd be much more to it.
Our clean room build infrastructure has suited us well for almost a decade now, providing deb and rpm repos and deb-packaged win32 and win64 crossbuilds for every bit of software (libraries, frameworks, apps, cross compilers, ...) we produce or need to reproduce.
Adding source repo support to the underlying software ("PBA") would require a lot of developer time and some valuable disc space, for no real gain whatsoever, so to me it'd feel like a waste of resources.
> I am one of those that utterly refuse to install a package (deb or rpm) if the source package is not available,
Suit yourself.
> If one want to make changes or customize the deb package,
... then one should and would use the upstream git anyway (which is fully debianzed) or maybe use the source packages from Debian or a Debian derivate.
I admit there are situations where source packages can be useful (to build upon some distro's integration and security patches), but this is no such case at all.
I haven't made any real changes, apart from unavoidable adaptions to our slightly non-standard and potentially over-customized build infrastructure.
There is one enhancement on my wishlist though that I might apply someday: The current packages enabling the apache2 conf snippet globally and without asking during postinst, that's a bad idea in the ages of https-everywhere, letsencrypt and SNI.
Fortunately our servers' /etc directories are "gittified" (under git control) now ;-))
Regards,
Christoph
|