[Agda] Warning: hashable-1.2.0.x has serious performance bugs

[Andreas Abel]

This is not a constraint at all since unordered-containers exists only 
up to 0.2.3.x, as far as I can see.

Probably you meant to constrain hashable < 1.2.  But if you have 
packages such as hashtables linked with 1.2.x-versions of hashable, 
cabal install might just fail, wouldn't it?

Also hiding packages does not prevent them to be linked into your 
program, so I rather be on the careful side and do proper unregistering 
of bad packages.

Cheers,
Andreas

On 12.07.13 8:48 PM, Felipe Almeida Lessa wrote:
> Doesn't it suffice to use
>
>    cabal install --constraint="unordered-containers < 1.2"
>
> ?
>
> Cheers, =)
>
> On Fri, Jul 12, 2013 at 3:16 PM, Andreas Abel <andreas.abel at ifi.lmu.de> wrote:
>> Hi Martin,
>>
>> I think that something like the following should work (assuming the darcs
>> version 2.3.3 of Agda, should be similar for 2.3.2 or 2.3.2.1):
>>
>>    ghc-pkg unregister Agda-2.3.3
>>    ghc-pkg unregister unordered-containers
>>    ghc-pkg unregister hashtables
>>    ghc-pkg unregister hashable
>>
>> Some of these steps may fail because there are other packages depending on
>> hashable/hashtable/unordered-containers.  In this case, either use
>>
>>    --force
>>
>> or/and also unregister and then reinstall the broken packages.
>>
>> Use
>>
>>    ghc-pkg list
>>
>> to ensure there are no hashable-1.2.x.y installations left.
>>
>>    cabal install hashable-1.1.2.5
>>    cabal install hashtables
>>    cabal install unordered-containers
>>
>> In the Agda directory
>>
>>    cabal clean
>>    cabal configure -v
>>
>> Check that no harmful version of hashable is used indeed.  Finally
>>
>>    cabal install
>>
>> Now
>>
>>    make library-test
>>
>> should run in a couple of minutes (4 on my medium-old laptop).
>>
>> You can use ghc-pkg list at any time to check the status of your package
>> collection.
>>
>> Hope this helps,
>>
>> Andreas
>>
>>
>> On 12.07.2013 19:09, Martin Escardo wrote:
>>>
>>> This explains a question I asked privately to you before, where I
>>> reported that Agda in a high-end desktop made this year ran at the same
>>> speed as in a low-end netbook made in 2008.
>>
>>
>> I just came about the problem yesterday, and it took my whole day to find
>> the source of evil and expell it.  Sorry for being clueless before, I used
>> to trust people to only upload extensively tested libraries onto hackage,
>> but now I learned the better.
>>
>>
>>> Question: can you be more precise regarding what we have to do to avoid
>>> this problem? (This is not urgent: I made the desktop run 9x faster by
>>> copying the .cabal tree from the netbook to the desktop, as I told you.)
>>>
>>> Thanks,
>>> Martin
>>>
>>> On 12/07/13 16:52, Andreas Abel wrote:
>>>>
>>>> When you compile Agda, make sure it uses hashable-1.1.2.x and not the
>>>> newer 1.2.0.x version (which cabal automatically pulls if the choice is
>>>> not constrained).
>>>>
>>>> With hashable-1.2.0.x, Agda is 10x slower!
>>>>
>>>> The newest darcs version asks for hashable < 1.2, so
>>>>
>>>>     cabal install
>>>>
>>>> might not work out of the box, you might have to clean out newer
>>>> versions of hashable and reinstall packages that used hashable-1.2.0.x,
>>>> like the hashtables package.
>>>>
>>>>     cabal configure -v
>>>>
>>>> should show you which version of the packages are picked for compilation.
>>>>
>>>> Cheers,
>>>> Andreas
>>>>
>>>
>>
>>
>> --
>> Andreas Abel  <><      Du bist der geliebte Mensch.
>>
>> Theoretical Computer Science, University of Munich
>> Oettingenstr. 67, D-80538 Munich, GERMANY
>>
>> andreas.abel at ifi.lmu.de
>> http://www2.tcs.ifi.lmu.de/~abel/
>> _______________________________________________
>> Agda mailing list
>> Agda at lists.chalmers.se
>> https://lists.chalmers.se/mailman/listinfo/agda
>
>
>

-- 
Andreas Abel  <><      Du bist der geliebte Mensch.

Theoretical Computer Science, University of Munich
Oettingenstr. 67, D-80538 Munich, GERMANY

andreas.abel at ifi.lmu.de
http://www2.tcs.ifi.lmu.de/~abel/