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

Re: [GSoC] blends-gen-control hints (Was: blends-dev, gsoc 2013)

Hello Andreas,

On Sun, Jul 7, 2013 at 8:26 PM, Andreas Tille <andreas@an3as.eu> wrote:

I also found a slight difference.  Not in all architectures but

  ./blend-gen-control --blend debian-med --architecture mips -c -D -S
  ./sec-blend-gen-control --blend debian-med --architecture mips -c -D -S

show a difference for tree-puzzle | tree-ppuzzle.

Thanks for mentioning, it was a bug in blend-gen-control, I fixed it.

OK, sounds good.  Just keep your idea alive for some time - there is no
reason to assume that something is correct just because it was my idea.

Of cource, but I don't vote for your idea just because it's yours :-). Let's take for example the above bug you mentioned for debian-med mips. sec-blend-gen-control had correctly added the "tree-puzzle | tree-ppuzzle" alternatives into suggests but blend-gen-control did not(it just added them as single packages, it was a bug in an if-else clause). This was the second time I found a bug in blend-gen-control using as reference the results of sec-blend-gen-control. The latter provides, as I said, a more solid handling of the alternatives. That because it starts from the initial alternative relation "pkg1 | pkg2 | pkg3 | pkg4", checks which packages are available or not and properly splits and moves the packages to the correct lists(depends, suggests etc), so it does not lose packages because it starts from the  alternative relation itself.

On the other hand blend-gen-control first fills in the lists(depends, suggests) with single packages and then using the alternatives it combines the single package. The first weak point is that in case you forget to combine the corresponding alternatives relation with the corresponding single list you will not get correct results.For example at first I was combining the Depends-single-list with the Depends-alternatives, the Suggests with the Suggest-alternatives etc.Let's say you have the alternatives Depends: "pkg1 | pkg2 | pkg3"  but because the pk1,pkg2,pkg3 are not available they will move into the Suggests list. So if you do not combine the Depends-alternatives with the Suggests-single-list, the packages pkg1,pkg2,pkg3 will stay as single elements in the Suggests' list. That was the first simple bug in blend-gen-control I found out using as reference the results of sec-blend-gen-control.

Anyway of course I will keep both ideas alive, I will make more tests today (and check if more bugs come to life). And as soon as we are sure about the results then we can we can decide about the final structure of UDD blends' tables and blend-gen-control

PS: I hope you had a nice trip travelling from Porto Alegre ;-)

Kind regards


Reply to: