[darcs-users] MinGW running problems

Gour gour at mail.inet.hr
Sun Nov 16 21:38:20 UTC 2003


David Roundy (droundy at abridgegame.org) wrote:

> What he means is that you can (as a workaround) modify writeToFile in
> Lock.lhs to removeFile f before renameFile newf f.  This has the
> unfortunate result that it makes repository writes non-atomic (if you lose
> power between the removeFile and the renameFile, your repo is hosed).

Probably he does not know that I didn't even start to learn Haskell :-)

> 
> You'd also want to make the same change in Zlib.lhs to gzWriteToFile (which
> serves the same purpose as writeToFile, except that it compresses it.

I'll take a look. Keep you fingers crossed :-)

> I would say that renameFile is very broken on released win32 ghc if it
> doesn't support renaming to an existing file.

Yes, we can safely coem to such conclusion, at least.

> Perhaps we should create an autoconf test for a broken renameFile? It
> might be a good idea to start switching to autoconf-style tests rather than
> "#if __GLASGOW_HASKELL__ < 600" for dealing with ghc bugs and old library
> versions.

imho, the 6.0.1 is not so brand new, and it looks that upgrade is not so 
expensive, so why bother to keep patching old compiler versions and wasting
valuable development time, especially in the present situation when darcs itself is dynamically moving target.

Later, when darcs reaches some stability, it is reasonable to try to keep it
compatible with some older ghc version.

Sincerely,
Gour

-- 
Gour
gour at mail.inet.hr
Registered Linux User #278493





More information about the darcs-users mailing list