[darcs-users] [Haskell-cafe] ANN: darcs 2.3 beta 1

[Petr Rockai]

Duncan Coutts <duncan.coutts at worc.ox.ac.uk> writes:
> Configuring HTTP-3001.1.5...
> Preprocessing library HTTP-3001.1.5...
> Building HTTP-3001.1.5...
>
> Why are we using HTTP-300x.x in new releases? In cabal-install-0.6.2 we
> switched over to HTTP-4000.x and it's great. HTTP-4000.x is used in the
> current Haskell Platform releases. It's faster and handles proxies
> better. I am aware that there are still some issues but I'm not aware of
> any regressions compared to the 300x series. The actual code changes to
> update to the new API are minimal. Are there particular reasons for
> sticking with the older versions?
No idea, probably because the original dependency was tight and no-one updated
it. Eric, is it OK to relax the depend?

> Building mmap-0.2...
> [1 of 1] Compiling System.IO.MMap   ( System/IO/MMap.hs,
> dist/build/System/IO/MMap.o )
> cbits/posix.c: In function ‘system_io_mmap_file_open’:
>
> cbits/posix.c:33:0:
>      warning: cast to pointer from integer of different size
> cbits/posix.c: In function ‘system_io_mmap_file_close’:
>
> (and 5 others similarly)
>
> This makes me very nervous. This is on Linux x86-64. Can someone look
> into that and fix it?
This is harmless, since the pointer is only ever used to store a file
descriptor. This is probably due to the way Win32 handles things, using
pointers to describe open files and the desire to keep the C API uniform over
POSIX/Win32 (the file handle is passed to/from Haskell).

There's a different bug in mmap that I have discovered, that's also easy to
fix, but I accidentally deleted the source, so I'll have to re-do that and send
patch upstream.

Yours,
   Petr.

PS: Thanks a lot for testing this.

-- 
Petr Ročkai | http://web.mornfall.net
A physicist is an atom's way of knowing about atoms. (George Wald)