[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Debian GNU/Hurd crossinstaller: "crosshurd"



Marcus Brinkmann <marcus.brinkmann@ruhr-uni-bochum.de> writes:

> On Fri, Jul 18, 2003 at 09:46:23PM +0200, Marco Gerards wrote:
> > Jeff Bailey <jbailey@nisa.net> writes:
> > 
> > > > Another problem right now is the translator vs. tarring up /dev stuff.
> > > > What are you going to do about this? With Marco's patch floating
> > > > around, wouldn't it make sense to talk to the Debian tar maintainer
> > > > (Bdale) while he's around? Or are we going to take a different road?
> > > 
> > > Different path.  It turns out that the only reason they don't use
> > > MAKEDEV is because it was slow.  We don't have that issue so will just
> > > use MAKEDEV.  I lost track of the thread for Marco's tar patch.  I don't
> > > remember if we ever decided it was an attribute or a file type.
> > 
> > (FYI:) It = passive translator
> > 
> > I was trying to explain the star maintainer that passive translators
> > are a filetype and not an attribute of a file. But he isn't convinced
> > of this yet...
> 
> That is because he is probably correct.  Remember that you can set a
> passive translator on any node, an empty file, a file with content, or
> a directory, or whatever.  On the other hand, it is not possible to
> set a translator to a symlink (I don't think you can stack passive
> translators, can you?)

I thought the contents of the file/directory were lost after setting a
translator on it, but I was wrong (I misunderstood the ext2fs code,
just by checking it out manually I found out you were right :)). But
that doesn't really matter, my patch makes a passive translator behave
like a filetype because that is the only data that is stored. Do you agree?
 
> So, it might very well be that the right behaviour for tar would be to
> tar up the underlying file and its translator setting (as an attribute
> to the file).  So far we only have ocnsidered empty nodes carrying
> passive translators.
> 
> > What matter more to me is having this patch (or a reworked version) in
> > GNU tar. This cannot be done in a POSIX compatible way without doing a
> > _LOT_ of GNU tar work. Jeff, what would be the best way to get this in
> > GNU tar? (Or at least to get some comments about this?).
> 
> For GNU tar, having a patch that just saves the translator setting
> (not the underlying file) would be a good first thing to have.
> However, saving the underlying file should also be consdered as an
> option, because some translators might require it in the future.
> POSIX compatibility is not a requirement for GNU tra today (as GNU tar
> is already not POSIX compatible).

I agree. When there is a new GNU tar (or the current one was modified
to be POSIX compatible) I can write a new patch. I'm aware that this
is only a short term solution.
 
> What would be more interesting is translator support in Linux ext2fs
> (to the extend of being able to read and write passive translator
> settings), so you could unpack such a tar file under GNU/Linux on a
> Hurd partition.

I agree. I thought about this. It is even possible to write a utility
that can do this on unmounted/read-only filesystems. (Just like there
is this program that can change the filesystem owner)

Thanks,
Marco



Reply to: