hello Aur: :) You get the point, the author of that package said that he has no enough time to check for the problem. then I have to solve it myself. For these kind of packages, maybe the best way to use it is re-write it myself... I have see a lot of packages that comes from "master thesis" or some. if you miss the season, then you miss the successful build :). 在 2011年10月15日 下午6:48，Aur Saraf 写道：

Did you at least receive this? :-)

---------- Forwarded message ---------- From: Date: Fri, Oct 14, 2011 at 10:47 PM Subject: Re: [Haskell-cafe] Cabal's dependence mechanism can be more available To: sonoflilit@gmail.com

You are not allowed to post to this mailing list, and your message has been automatically rejected. If you think that your messages are being rejected in error, contact the mailing list owner at libraries-owner@haskell.org.

---------- Forwarded message ---------- From: Aur Saraf To: 吴兴博 Date: Fri, 14 Oct 2011 22:47:27 +0200 Subject: Re: [Haskell-cafe] Cabal's dependence mechanism can be more available That a package builds with a certain version of another package does not mean that the code is correct.

I once had to maintain code that used "CSV > 0.0.0.1", and it turned out that in CSV-0.0.0.2 there wasn't a newline appended to the end of the string anymore (in the old version there was), so the packages compiled great together but I had a huge bug (because I was generating a text file of empty-line separated CSV blocks for a perl script to turn into an Excell file, and the perl now thought that it was all one big CSV block).

Trusting the author is the best that can be done in these cases. Sometimes the author is wrong, and then we must ask him to correct his packages - which I do a lot. But that's life.

Cheers, Aur

...

[[ My point is on cabal, but I had to give some examples. I'm very sorry for the authors of the packages listed below. ]] The story is : I was trying to install a package from hackage [holumbus-mapreduce]. I use ghc-7.2.1 and had a hard time to install the packages that depends on. [holumbus-mapreduce] depends on [holumbus-distribution]. [holumbus-distribution] depends on [hxt] and [hxt-regex-xmlschema]. etc.. There are tens of depends that drive me into an infinite 'step in'. If the dependence mechanism of cabal packages is good enough and the author of the package built it, the success can be repeat. but it not. a example: 1. There are hxt packages with version 9.* in hackage now. [holumbus-distribution] says that it depends on "hxt >= 8.2". But I had to add "< 9" to make it compiled. Maybe when holumbus-distribution was released, there was no hxt 9.*. 2. some packages depends on "haskell98 <2" I just removed this line, and it compiled, and id I use "haskell98", the newer "haskell98-2.*" will clash with new "base" 3. here is a piece of message from `runhaskell Setup configure` """ Configuring Holumbus-Distribution-0.1.1... Warning: This package indirectly depends on multiple versions of

On Thu, Oct 13, 2011 at 4:05 PM, 吴兴博 wrote: the

...

same package readline-1.0.1.0 requires process-1.0.1.5 package hslogger-1.1.5 requires process-1.0.1.5 package haskell98-2.0.0.0 requires process-1.1.0.0 """ but finally I found that "readline" does not refuse to be compiled with newer "process" at all, I use --reinstall to adjust "readline" and "hslogger" play with "process-1.1.0.0" It is like C++'s multi-deriving or something like that, I'm not sure of the technique word. A what to contain B,C ; B contains D' and C contains D'' ; but there can be only One D in A.

All I did is download the all packages, edit its depends: and compile and install it.

If Cabal can remember some certain package-version combination of the successful builds, Sure It will make hackage more available to users. based on some success build, it can also try to make promotion to depends-packages, and provides newest & can build packages.

Cheers!

--

Wu Xingbo

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

-- 吴兴博 Wu Xingbo