Re: hackage importer

[Ludovic Courtès]

Federico Beffa <address@hidden> skribis:

> On Thu, Mar 26, 2015 at 2:09 PM, Ludovic Courtès <address@hidden> wrote:
>> Could you post the actual backtrace you get (?) when running the program
>> with LC_ALL=C?
>
> I doesn't backtrace, the function just gives the wrong result.

Hmm, OK.  Still sounds like an encoding error.

>>>> I would expect the conversion of conditional expressions to sexps to be
>>>> done in the parsing phase above, such that ‘read-cabal’ returns an
>>>> object with some sort of an AST for those conditionals.
>>>>
>>>> Then this part would focus on the evaluation of those conditionals,
>>>> like:
>>>>
>>>>   ;; Evaluate the conditionals in CABAL according to FLAGS.  Return an
>>>>   ;; evaluated Cabal object.
>>>>   (eval-cabal cabal flags)
>>>>
>>>> WDYT?
>
> I'm not sure this can be done, because flags must be declared in the
> cabal file itself and the manual is not clear if the flags are
> required to be declared before they are used or not (although it would
> make sense).

By ‘flags’ I actually meant things like ‘unix’, which appear in
conditionals:

  (eval-cabal cabal '(unix))

> To see how this importer (and the haskell-build-system that I've
> posted) performs I've now used it with several libraries (for the
> moment I've packaged 26 libraries). While it runs nicely with most of
> them, a couple posed problems resulting in package definitions with
> some fields empty and 1 caused a backtrace. (Once I fixed the package
> manually, the build-system worked fine for all of them.)
>
> Before working further on improving the interface, I want first to
> understand what are the root causes of the errors (especially the one
> causing the backtrace) and fix them.

Sounds good, thanks!

Ludo’.