[darcs-devel] Re: darcs-unstable fails to compile

[Mark Stosberg]

On 2005-04-02, David Roundy <droundy at abridgegame.org> wrote:
> On Sat, Apr 02, 2005 at 01:59:24AM +0000, Mark Stosberg wrote:
>> Hello,
>> 
>> This is what happens when I try to compile darcs-unstable on
>> Mandrakelinux with ghc 6.4:
>> 
>>  $ make
>>  ghc  -cpp  -package parsec -package unix -O -funbox-strict-fields  -Wall -Werror -package util -I. -c Record.lhs
>>  
>>  Record.lhs:105:4: Not in scope: `withLock'
>>  Record.lhs:124:40: Not in scope: `existing_files'
>>  Record.lhs:320:22: Not in scope: `removeFile'
>> 
>> ####
>> 
>> Any hints? 
>
> It's a weird problem.  My darcs-unstable doesn't have those functions on
> those lines, so perhaps somehow you've got some repository corruption, or
> perhaps you've got an older version of darcs-unstable?
>
> $ md5sum Record.lhs
> e43a7fbca452b92c06d692976374cb9e  Record.lhs
>
> You could try running darcs check, perhaps.  Needless to say, I don't see
> this behavior.  It's odd, because existing_files is a local variable.  And
> withLock has been replaced by withRepoLock, so my Record.lhs doesn't have
> withLock in it at all.  :(
>
> I'm in the process of getting a fresh copy, in case somehow things get
> broken in transit, but I'm guessing that somehow something went wrong in
> your copy.  Do whatsnew or resolve show anything?

Resolve shows something 'interesting':

  Fail: bug in darcs!
  Problem resolving conflicts in resolve!
  Please report this to bugs at darcs.net,
  If possible include the output of 'darcs --exact-version'.

That's from Mandrake's 1.0.2 binary. 

'darcs check' thinks the repo is consistent.

Although my memory is not strong, It's possible I got a 'conflict' in
that file while updating recently, and resolved it with a 'revert'. I
figured since I hadn't made code modifications to that file myself, that 
this was a reasonable solution. Darcs didn't complain. :)

Maybe that was the wrong thing to do, but since I'm not a Haskell
programmer, guessing which piece of code was correct wouldn't have been
any better.

Maybe if 'resolve' worked, it would re-introduce that conflict for me? 

    Mark


-- 
http://mark.stosberg.com/