[Haskell-cafe] Plugin Problem - Weirder

[Donald Bruce Stewart]

daniel.is.fischer:
> Am Donnerstag, 21. Juni 2007 10:22 schrieb Bayley, Alistair:
> > > [mailto:haskell-cafe-bounces at haskell.org] On Behalf Of Daniel Fischer
> > >
> > > I suppose in contrast to the version from HackageDB, which I
> > > got myself on
> > > monday, the darcs version works with ghc >= 6.6, or I
> > > probably would have
> > > heard about it. So I'll try to get me that, only where?
> >
> > http://www.cse.unsw.edu.au/~dons/hs-plugins/
> 
> I'd already tried that, but
> $ darcs get --partial http://www.cse.unsw.edu.au/~dons/hs-plugins/
> Invalid repository:  http://www.cse.unsw.edu.au/~dons/hs-plugins
> 
> darcs failed:  failed to fetch: 
> http://www.cse.unsw.edu.au/~dons/hs-plugins/_darcs/inventory ExitFailure 1

The file system was down here, sorry.  Should be up now.

    $ darcs get http://www.cse.unsw.edu.au/~dons/code/hs-plugins/

> Is my darcs too old (version 1.0.5) or is the repo actually broken?
> >
> > AFAIUI, version 1.0 will work with ghc6.6 on non-Windows platforms. It's
> > been broken on Windows for some time, and I don't really know why. As
> > you seem to be on Linux, this shouldn't affect you.
> >
> > Alistair
> 
> Well, again, the .tar.gz from HackageDB is tagged 1.0, it definitely doesn't 
> work with ghc-6.6.1 on my linux. And it apparently was packed in February, so 
> it shouldn't be too out of date.
> 
> Let me summarize my findings so you guys have something to reason from.
> 1) if I compile the module and execute the binary, I get
> $ PlugTest
> PlugTest: /tmp/MKtHcN2822.o: unknown symbol 
> `pluginszm1zi0_AltDataziDynamic_toDyn_closure'
> PlugTest: user error (resolvedObjs failed.)

You must use the darcs version with ghc > 6.6

> 
> 2) If I load the compiled code into ghci, load cannot ever find the symbol 
> resource, originally that is because it looks for main_resource_closure 
> instead of ModuleName_resource_closure.
> Now I changed the code in System.Plugins.Load, so that load does look for the 
> correct symbol, with baffling results. In compiled code, all loads still 
> fail. If I load interpreted code into ghci, the first load succeeds, hooray! 
> But all subsequent loads fail, until I :reload (sometimes) or :load PlugTest 
> (that always worked so far). How come? When I added debugging output, I could 
> see that it looked for the correct symbols in the correct .o files, but only 
> the first found the symbol. Very peculiar.
> 
> Could somebody please enlighten me?

The module name issue is fixed in the darcs version.

-- Don