From 1be510f9529cb082f802408b472a77d074b394c0 Mon Sep 17 00:00:00 2001 From: Nicolas Vigier Date: Sun, 14 Apr 2013 13:46:12 +0000 Subject: Add zarb MLs html archives --- zarb-ml/mageia-dev/20110315/003345.html | 157 ++++++++++++++++++++++++++++++++ 1 file changed, 157 insertions(+) create mode 100644 zarb-ml/mageia-dev/20110315/003345.html (limited to 'zarb-ml/mageia-dev/20110315/003345.html') diff --git a/zarb-ml/mageia-dev/20110315/003345.html b/zarb-ml/mageia-dev/20110315/003345.html new file mode 100644 index 000000000..b4fd35058 --- /dev/null +++ b/zarb-ml/mageia-dev/20110315/003345.html @@ -0,0 +1,157 @@ + + + + [Mageia-dev] Packaging errors to fix + + + + + + + + + +

[Mageia-dev] Packaging errors to fix

+ Maarten Vanraes + maarten.vanraes at gmail.com +
+ Tue Mar 15 22:57:59 CET 2011 +

+
+ +
Op dinsdag 15 maart 2011 19:06:28 schreef Michael Scherer:
+> Le mardi 15 mars 2011 à 18:34 +0200, Anssi Hannula a écrit :
+> > On 15.03.2011 15:38, Michael Scherer wrote:
+> > > 3rd : dbus. There is a vicious loop between dbus and its own library,
+> > 
+> > So? Both of the requires are correct from what I know, and dependency
+> > loops are not forbidden.
+> > 
+> > I can imagine there are also some more clear cases of such loops, like
+> > library A depending on B, B on C, and C again on A, which I personally
+> > don't believe are in the need of "solving".
+> 
+> We did have some troubles with loops at building time when we
+> bootstrapped the distribution, and we clearly wanted to remove them as
+> much as possible.
+> 
+> A loop at runtime is potentially the sign of a loop at build time too.
+> So we should try to avoid them IMHO.
+> 
+> In the case of dbus, yes, the requires are correct, but the ordering
+> seems wrong during the test. So either there is a bug in what ever jbj
+> is running, or we do use Requires(post) in a wrong way, or that's a
+> change somewhere. In any case, there is still the others problem to look
+> at :)
+
+imho, i think if loops are unneeded or solvable, we should try to do that if 
+it's doable in a short time.
+
+when we were talking about loops at bootstrap time, this is a whole different 
+area, i think most of the loops are not really solvable, but we should find a 
+way to make it easier and possibly documented, or even automagically...
+
+for example:
+ - you can bootstrap X by installing Y without functionality Z
+
+if we would have some way to note how this is done, perhaps we can even do it 
+automagically on buildsystem level.
+
+looking at this from my pov, perhaps having a spec file macro isn't a bad idea:
+
+%bootstrap(%feature1 %feature2 ...)
+
+would be nice, if building it with for example --bootstrap, would deactivate 
+all those features
+
+then you could let buildsystem build once with --bootstrap and once without it 
+after the dependencies are done
+
+if you'd look at things even more, we could do something like this:
+
+%build_order(<group>:<order_nr>) 
+
+then with buildsystem one could build a bunch of packages and the buildsystem 
+could order them according to the group.
+
+
+eg java: (totally fictional example, i know nothing about java packaging)
+============================================
+
+in package A:
+----------------
+%bootstrap(maven eclipse)
+%build_order(java:1)
+
+%if %maven
+...
+%endif
+
+
+
+in package B:
+----------------
+%bootstrap(maven)
+%build_order(java:2)
+
+%if %maven
+...
+%endif
+
+
+in package maven:
+-----------------------
+%bootstrap(eclipse)
+%build_order(java:500)
+
+
+
+you will all agree that i'm talking bollocks here, as this is totally not how 
+it is.
+
+However, if you have something like this, you could theoretically say to 
+buildsystem: "build java for me"
+
+it could then find all spec files with build_order java, sort them on the 
+number, and build and submit them all in the order. and after it's done, do 
+this again with the ones that had bootstrap defined.
+
+
+to me, it sounds like a good theoretical solution and it would still be 
+usefull, even after bootstrapping. no idea if it would take alot of work or 
+not or even practically possible.
+
+ + + +
+

+ +
+More information about the Mageia-dev +mailing list
+ -- cgit v1.2.1