_sopen (was: Corrupted database)

Matt Oberle wrote:
 > If we had spent all this time and energy working together to implement
 > Pat's solution this would be a moot point.

by the way, does someone knows about _sopen() ?

-------- Original Message --------
Subject: Re: [Metakit] Corrupted database over network - WinNT
Date: Wed, 14 Feb 2007 16:09:13 +0100
From: Riccardo Cohen <[email protected]>
Organization: articque
To: [email protected]

Hi Jean-Claude,
This is very clear, thank you. I really hope that vista corrected this wi=
ndows bug. It is
not the only bug with windows networking, but this one is a real pain.
The problem now for us at least, is that we have many different way of us=
ing metakit.
sometimes we encapsulated in one of our class, wich make it easy to redef=
ine a file
strategy, but sometimes we access directly for performance reason, and th=
ere are many
places where we open a storage.

Maybe I'm wrong, but there might be a very easy and effortless solution (=
for developers
that never have 2 C4_storage on the same file) :

in c4_FileStrategy::DataOpen(), I tried to replace :

fd =3D _wopen(wName, flags);

by :

fd=3D_wsopen(wName,flags,_SH_DENYRW);

Unfortunately it did not work, but really if someone know something about=
 windows open()
function, this could help many people.

Thanks for any clue :)



Jean-Claude Wippler wrote:
> Thanks, Pat Thoyts for your detailed analysis and description of the=20
> workaround you found.
>=20
> Let me summarize as follows:
>=20
>> Metakit maps the file as
>> a read-only memory mapping and when it saves it re-serializes into the
>> underlying file. It then drops the current mapping and re-maps the
>> file. What I observed was that when it re-mapped it got the original
>> data and did not in fact re-map the newly written data at all.
>=20
> In other words, Windows does not guarantee a consistent view of the dat=
a=20
> it buffers for networked file systems: write something to a file, map i=
t=20
> to memory *afterwards*, and apparently you can get stale data.
>=20
> If MK were to use r/w maps and push its data through there, then=20
> presumably it would work.  That will not work in the case when the file=
=20
> has to be extended, though.  Keep in mind that MK can write data to fil=
e=20
> for which partially comes from the r/o map.  Furthermore, MK maps r/o=20
> deliberately, because it prevents stray pointer writes from damaging a=20
> MK datafile.
>=20
> So what we have today, is a design which works on Windows local disks,=20
> and on Unix/MacOSX both locally and on remote file systems (both NFS an=
d=20
> SMB afaik).
>=20
> [From: <[email protected]>]
>> 1/ You confirm that metakit files have a serious design flaw that puts=
=20
>> at risk our data when accessed through a network.
>=20
> With all due respect, the "serious design flaw" is in Windows.  Getting=
=20
> that fixed is beyond our control.  We can only hope that XP or Vista ge=
t=20
> it right.
>=20
>> 2/ You give indications on the nature of the corruption, and how to=20
>> prevent it.
>=20
> Yes, and for that I am very grateful.  In addition to Pat's most recent=
=20
> post, more information can be found at=20
> http://www.equi4.com/pipermail/metakit/2005-August/002112.html
>=20
> So the solution is to derive a c4_Strategy class which opens the file i=
n=20
> exclusive mode.  Pat has most generously handed everyone the solution o=
n=20
> a silver platter.  If you don't want to subclass c4_Strategy, then open=
=20
> the file yourself and pass it to MK.  If you never need concurrent=20
> readers, use CreateFile and always open with dwAccessMode set to 0, the=
n=20
> turn it into a FILE* with freopen and pass it to MK.
>=20
> I for one never use MK in r/w mode on remote file systems.  Not just du=
e=20
> to the above risks, but because locking also leads to all sorts of=20
> "Heisenberg" trouble.  In my book, shared data use should use=20
> client/server solutions.  That need only take a few dozen lines of code=
,=20
> depending on your tool choices.
>=20
> -jcw
>=20
>=20
> _____________________________________________
> Metakit mailing list  -  [email protected]
> http://www.equi4.com/mailman/listinfo/metakit
>=20

--=20
Tr=E8s cordialement,

Riccardo Cohen
ligne directe : 02-47-49-63-24
-------------------------------------------
Articque
http://www.articque.com
Moulin des Roches
37230 Fondettes - France
tel : 02-47-49-90-49
fax : 02-47-49-91-49
_____________________________________________
Metakit mailing list  -  [email protected]
http://www.equi4.com/mailman/listinfo/metakit


--=20
Tr=E8s cordialement,

Riccardo Cohen
ligne directe : 02-47-49-63-24
-------------------------------------------
Articque
http://www.articque.com
Moulin des Roches
37230 Fondettes - France
tel : 02-47-49-90-49
fax : 02-47-49-91-49
_____________________________________________
Metakit mailing list  -  [email protected]
http://www.equi4.com/mailman/listinfo/metakit