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

Re: [MiNT] MiNT and MetaDOS


Sorry for replying late, but this message was already POPped to my home
computer, which I rarely use these days.

On Wed, May 03, 2000 at 01:08:00AM +0300, Martin-Éric Racine wrote:
> If it was only SPIN!, the latest FNRAMFS would not freeze the
> kernel dead while booting, if used on kernels prior to 1.15.7,
> would it?

Frank already commented on that.

> That does not get him any more answers than it does for me

So, you think rude accusations are the way to go? Get a life ...

> does it get the message that changes to _documented_ features are
> meant to be _discussed_ and _agreed_ on this list.

Again, you're making statements without knowing the facts. The actual
change (*), as far as Frank has told me and here on the list, was a
semaphore check to prevent an XFS's function to be called while another
one of this very XFS is still active - something like this is necessary
to get useful background DMA, as you can't assume all existing
filesystems are reentrant. (New filesystems can set a flag to tell MiNT
they are and therefore do not need that semaphore locking.)

The problem with SPIN! now seems to have been that it internally calls
one of its own functions in certain situations, causing a deadlock for
the process that's initiated the access. Frank and/or Julian, please
correct me if that's wrong or inaccurate.

Reading the XFS documentation, I can't find any passage that states
"calling of XFS functions within an XFS function is allowed". In fact,
there's no comment on XFS reentrancy there, as it wasn't an issue when
the text was written. So I can't consider blocking such events "breaking
documented behaviour" at all. I wouldn't even have thought it could
break an existing XFS had I implemented that semaphore locking.

> If everything is gonna be decided between Frank, yourself and Guido,
> we might as well discontinue this mailing list...

You over-estimate my involvement in kernel development for sure. Aside
from that, the main problem with discussing things here before doing
something is that up to now posting a proposal has quite often just led
to lengthy threads with no real result - something I wouldn't consider
speeding up the development process, if done for each and every change
that might have some negative effects ...

Besides, beta versions (and the problem with SPIN! arose in a beta
version) are there to be tested - not to be used as a simple replacement
for the current release version. If something doesn't work with a beta,
report it. And if someone reports a problem, please inform yourself
about the background before yelling about bad, bad breaking changes.

EOT from my side.



(*) The code for this is already in the source for several versions
    (since 1.15.3b), but up to 1.15.7b, it hasn't been compiled in.

Thomas Binder (Gryf @ IRCNet)  gryf@hrzpub.tu-darmstadt.de
PGP-key available on request!  binder@rbg.informatik.tu-darmstadt.de

Attachment: pgpYBG2bJfxHn.pgp
Description: PGP signature