Re: Tagged paging area?

[Roland McGrath <roland@frob.com>]

> Is anyone changing the pager to require a SWAP-SPACE tag on the
> partition, as in Linux?  With the current scheme, I fear I could
> lose a filesystem if I remove a logical partition and forget to
> tell GNU.

We are adding code to recognize Linux swap signature pages when they are
there and obey the format they specify (i.e. obey their bad-block maps, and
don't clobber the signature page itself on disk).  

But you seem to want something different, that we refuse to swap on a
partition without a linux signature page.  We are certainly not going to
make that the default, since we believe in swapping to specified whole
partitions or files as other systems do, without requiring any special
preparation of the scratch data.  

What I will do is add an option in the boot script to check for the signature.
Right now you add initial paging files (or partitions) with a line in the
boot scripts (/boot/servers.boot) like:

	/dev/hd0s2 $(add-paging-file) $(default-pager)

With the new option you will do:

	/dev/hd0s2 $(add-linux-paging-file) $(default-pager)

In both cases, if there is a linux signature page, the pager will read and
obey it (it prints a message on the console if it sees a signature).  But
with $(add-linux-paging-file), the pager will print an error message and
skip the file if it has no signature.  (While I'm at it, I think I'll add
an $(add-raw-paging-file) that never checks for a signature, for people who
really want to clobber their whole partitions no matter what.)

How does that sound?

> Or should I just convert to disklabel?

I don't understand how that's relevant.  But then I don't really understand
what your exact concern is.

> Meta-question:
> Is it okay to post such feature requests here, or should I submit
> wishlist-priority bugs instead?

Since this is about a hurd feature, not a debian feature, the general
answer is that you should post it to bug-hurd@gnu.org instead, because that
is the canonical way to get to the hurd maintainers, who do not necessarily
read debian-hurd (notwithstanding the fact that at the moment we all in
fact do).  OTOH, things are now set up so that debian-submitted bugs for
the hurd package get sent to bug-hurd@gnu.org; and that way it will stick
around in the debian BTS and perhaps be less likely to be forgotten than if
we just get the message on the bug list.