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

Re: mass-commit: dependencies



On 12-07-06 at 04:09pm, Dominique Dumont wrote:
> On Friday 06 July 2012 15:30:23 Jonas Smedegaard wrote:
> > On 12-07-06 at 10:35am, Dominique Dumont wrote:
> > > > d) is harmful: a recent cdbs _is_ needed.  In this particular 
> > > > case the CDBS snippet perl-makemaker.mk is included which was 
> > > > unavailable prior to cdbs 0.4.73.
> > > > 
> > > > It could be argued that d) is not harmful because only unstable 
> > > > (and testing) is targeted - backporting is unsupported 
> > > > officially.  In that case this particular versioned dependency 
> > > > is indeed unneded - but then so are versioned build-dependencies 
> > > > on debhelper even in compatibility level 8 and 9!
> > > 
> > > I would not recommend to remove debhelper > 9 as this would harm 
> > > backport on squeeze.
> > 
> > I don't understand above sentence.
> 
> Well, when backporting, you have to consider whether a package is 
> compatible or not with with your target. If you consider backporting a 
> pacakge with debhelper > 9 for squeeze, you know that you will have 
> some rework to do.
> 
> If the required dep is removed, people doing backports for squeeze 
> will have more err, unwelcome surprises.

These are the debhelper versions in the officially released suites:

Lenny: 7.0.15
Squeeze: 8.0.0.

If you are not taking backports.debian.org into account, I suspect you 
really mean debhelper > 8.


> > Please mention backports.debian.org explicitly if you are talking 
> > about backporting to there specifically.  I don't.
> 
> Neither do I. sorry. The main question when cleaning dependencies is 
> where to draw the line. IIRC, the consensus was to remove version req 
> that can be satisfied on oldstable. (which is admitedly a blurry 
> notion since wheezy froze).

Not at all blurry to me: It makes sense to support backporting to 
officially maintained releases of Debian, when reasonable.  Freeze does 
not affect this - cancellation of security support for oldstable does.

Versioned package relations satisfied in all officially maintained 
Debian releases can safely be simplified to unversioned relations.

This is the reason I drop versioning when using debhelper compat level 7 
(and I do not depend on any specific features introduced later than the 
major release of the tool).

When Lenny is no longer officially supported by Debian, I will use 
debhelper compatibility level 8 by default.

(and no, multiarch support does *not* require debhelper 9!)

I require recent CDBS more frequently than I require recent debhelper.  
Reason for that is that CDBS itself is is easy backportable to 
oldstable, unlike debhelper.



> > > Note that, by default cme will not remove versioned dependencies 
> > > required for oldstable, provided the information is provided by 
> > > madison.
> > 
> > I don't understand what you mean here.  How can madison know if what 
> > features of CDBS (or debhelper or whatever) used speficially in a 
> > certain package. causing a versioned dependency?
> 
> That's not the point. madison is used to know whether the versioned 
> dependency can be satisfied in all supported cases (usually taking 
> into account oldstable). If yes, the versioned req is removed.

I complained about two things regarding versioned dependencies:

 * unneeded debhelper versioning was added
 * needed cdbs versioning was dropped

Your comment above is about Config::Model relying on Madison for 
_removal_ of versioning, right?

Is your comment above somehow related to my complaint about dropping 
needed cdbs versioning?


 - Jonas

-- 
 * Jonas Smedegaard - idealist & Internet-arkitekt
 * Tlf.: +45 40843136  Website: http://dr.jones.dk/

 [x] quote me freely  [ ] ask before reusing  [ ] keep private

Attachment: signature.asc
Description: Digital signature


Reply to: