This is the mail archive of the automake@gnu.org mailing list for the automake project.


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

Re: Automake 1.4l released


Tim Van Holder wrote:
> 
> > the old original).  This is allowed on linux, but NOT on
> > windows/NTFS/cygwin, when perms = -r--r--r--.
           ^^^^^

> 
> I'm pretty sure this is a cygwin issue (I'd even consider it a
> bug in cygwin's behaviour, as it matches neither Unix nor
> Windows behaviour).  On DJGPP (in a DOS box under WinME, FAT32
> drive) I get:
> 
> The only potential problem with DOS/FAT/whatever access times
> in this regard is that directories can't have their timestamp
> set using the usual DOS system calls; the LFN API introduced
> with Win9x does offer a system call for that, but basic DOS
> does not allow it.
> 
> Also, Linux' filesystem drivers for fat/vfat/... don't support
> setting all file times, so they cause similar trouble.
> 
> For regular files, on a DOS/Windows based system, there should
> be NO such errors IMHO.

But I'm not talking about FAT or VFAT.  I'm talking about NTFS, a whole
different beast.  IMNSHO, the linux and dos behaviors you describe ARE
the mistake, and cygwin's behavior (on NTFS) is correct -- why should
you be allowed to change the file metadata (timestamp, etc) if you don't
have write access to the file?

That's why I advocate the change: what if other filesystems behave
"correctly" (my definition).  In addition, what is the rationale for
doing 'chmod -R a-w' in the distdir: rule, anyway?  Isn't 'chmod -R
go-w' just as "good" a choice?

--Chuck


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