This is the mail archive of the cygwin-developers@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]

Re: readonly, NTFS, and file metadata


Earnie Boyd wrote:


 > But, this only solves it for NTFS with ntsec set.  What about FAT?  I
 > think that it is wrong of automake maintainers to be so narrow minded.
 > I plan to join that discussion.

I finally got a response on *why* the distdir was chmod'ed '-R a-w' 
It's because the test is to see if a distribution (of the package you're 
automaking) will operate correctly if the sources are on a read-only 
medium.  Thus, preserve time stamps, but make sure EVERYTHING is read-only.

I'm not sure how this would interact on FAT drivers, or NTFS/nontsec. 
Its probable that 'cp -p' would just do its job and return without 
error, and automake would be none the wiser.  Maybe that's what cp -p 
should do -- just ignore the ENOACCESS that the SetFileTime() function 
returns?

One person on the automake list suggested:

 > Maybe I'm missing something, but my manual page for cp reads
 >
 >        -p, --preserve
 >               preserve file attributes if possible
 >
 > Doesn't `if possible' means cp should not complain is it does?

BTW, with NTFS/ntsec (Aug 11 snapshot), I can

mv foo bar

successfully, even if foo is -r--r--r-- (I own foo).

--Chuck



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