6.4.1 Bug? getServiceEntry: does not exist (no such service entry)

[Simon Marlow]

S. Alexander Jacobson wrote:

> Ok, I guess the easiest way to reproduce is:
> 
>   0. curl http://searchpath.org/searchpath/SearchPath.hs > SearchPath.hs
>   1. ghc --make -main-is SearchPath SearchPath.hs -o sp
>   2. darcs get http://pass.net/s/repo pass.net
>   3. cd pass.net
>   4. ..\sp ghci  -DINTERACTIVE -isrc --internet 
> http://searchpath.org/default.map src/Main.hs

I think you mean "src/PassNet.hs" here.

>   5. main
>   6. open firefox or IE to http://localhost:8000
> 
> Searchpath will take care of downloading all the required libraries into 
> the .haskell_cache directory.
> Normally hitting 'e' at this point should exit main and return you to 
> the ghci shell, but that doesn't appear to be working in 6.4.2 either or 
> responds so slowly as to effectiveyl be the same.

Ok, I don't really know what I'm looking for, but here's what I'm 
seeing.  On Linux, with 6.4.2, I can browse to localhost:8000, then I 
hit 'e' and get

----------
e<interactive>:


FAILEDthread killed
----------

but sometimes it does nothing at all.

If I compile the program with -threaded, then 'e' doesn't exit.  It 
seems one of the threads is blocked in a read() call, I don't know why 
this should be the case.

Can you possibly reduce the test case at all?  The problem seems to 
occur with -threaded (GHCi is compiled with -threaded in 6.4.2, which is 
why you would see a difference there).

On Windows I can't seem to get it to serve anything over localhost:8000, 
I don't know why.

Cheers,
	Simon



> Note: What appears to be happenning is that the whole thing is just 
> sloowwww under 6.4.2.  It appears to be doing the right thing, but at 
> such a slow pace as to be unusable.