Haskell Platform Proposal: HLint

[Simon Marlow]

On 11/11/2010 20:29, Duncan Coutts wrote:
> On 11 November 2010 14:50, Simon Marlow<marlowsd at gmail.com>  wrote:
>
>> But I vote against allowing LGPL libs in.  For a start, wouldn't it have to
>> be LGPL + static linking exception anyway?
>
> In principle no static linking exception is required. It is possible
> to ship code in a relinkable form.
>
> What you do is to compile your app to .o files, link them together
> into one big app.o file, then link this statically with the dependent
> packages to make your app.exe. Then you ship the app.exe and the app.o
> (or a promise to provide it). Any user can then relink the app.o to
> get the app.exe.
>
> I have done this and automated it with Cabal. It would probably be
> useful to other people to add this feature to Cabal directly (I
> implemented it in a Setup.hs).
>
> Obviously it is rather tricky to generate a modified library that has
> the same ABI but the LGPL makes allowance for that. The LGPL-2.1
> states: "It is understood that the user who changes the contents of
> definitions files in the Library will not necessarily be able to
> recompile the application to use the modified definitions.". Current
> GHC does put us at the rather extreme end of that interpretation. Of
> course there is the same ABI issue with shared libs.

Right, my main objection has always been that since GHC makes it 
impractical to do the modify/relink that the LGPL requires, LGPL Haskell 
code isn't really complying with the spirit of the LGPL here.

I get the impression that people who use the LGPL for Haskell don't care 
deeply about this - they're using the LGPL because they want 
modifications contributed back.  In which case, wouldn't it make sense 
to add an exception for the modify/relink too?  Are there Haskell 
authors out there using the LGPL who really do care about modify/relink?

Cheers,
	Simon