This is the mail archive of the cygwin-apps@cygwin.com mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: ITP: leafnode


alper wrote:
On Fri, Apr 02, 2004 at 04:17:16PM +0200, Corinna Vinschen wrote:

1. Leafnode relies on hard-link counts, so current implementation of hlinks on FAT32 breaks it. News spool has to be on NTFS, and it seems this will be an NT/2K/XP only package.

Is that necessary? Can't that be ifdef'd for Cygwin?




I've discussed this before with leafnode maintainer, Matthias Andree, and here's what he says:

<quote>
The whole /var/spool/news/* organization relies on hard links.

The expire process relies on the hard link count (st_nlink in struct
stat, as in the <sys/stat.h> header file and returned by the stat()
syscall) being correct, file copies on FAT32 will lead to premature expiry and a horrible waste of disk space, it may become unbearable for cross-posted articles.
</quote>



To diverge, and I know Corinna will tell me to submit a patch, but I think the problem this raises is worth discussing. Even if we are using NT/2k/XP/2k3, we still could be using fat32 filesystem. So, another approach to fix this problem would be to extend the "managed" mount api. That is, put Cygwin in charge of maintaining an internal accounting of hard links which is agnostic to the filesystem being used. U/Win has had this approach for years now, so we know its possible. Of course, it be nice if old man Korn would let us see the source ;-). Just a thought.


As for the package, I tend to agree with Corinna. I don't think we should be packaging nt-only packages unless absolutely necessary.

Cheers,
Nicholas


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]