...uhm...
I am using a CENTOS 5.3 and I suppose that kernel do not have bug... I
have updated system regularly.
Really I am a little confused....
Using my wrong "filebuf.c" (that do not use hb_fileFind()) I have tried
to run the following program.
function main()
set exclusive off
dbcreate("temp",{{"line","C",10,0}})
use temp new
append blank
replace line with "pippo"
close all
use temp new alias temp1
?temp1->(rlock()) // .t.
use temp new alias temp2
?temp2->(dbrlock()) // .t.
close // close temp2
select temp1
temp1->line:="test" // in POSIX lock system should I receive a
"Lock Required" error message ?
close
use temp new
?line // program show "test"
quit
I expected to see some error or something that doesn't work, but seems
that everything works perfectly using my Centos (Redhat).
What do you think about it ?
Marco
Przemyslaw Czerpak ha scritto:
On Mon, 20 Jul 2009, Marco Bernardi wrote:
Hi,
I think I was not clear enough and confused you so I'll try to be more
precise because it may be source of some serious problems.
I am using linux 2.6 and seems it is like you say, close() didn't release lock.
It for sure release the fcntl() locks as all other POSIX systems.
Otherwise it's very serious bug in your kernel and I suggest to
check why it does not work correctly. It's documented POSIX behavior.
I was writing about BSD locks which we use in Linux to emulate
DOS deny flags which are not removed by current Linux kernels
and also do not interact with POSIX locks we still need for all
other synchrinizations so we can safely use them.
But the only one place were BSD locks are usable for us is DENY
flags emulation and for all other things we still need POSIX locks.
I have done following test: I have modified "filebuf.c" blocking the
hb_findFile() function... so any use command create a new handle and
it should create problems in POSIX locks.
Yes, and it does.
I created a short programs in order to check it in Linux the programs give
net-error while I try to open TEMP in Exclusive: if I have well understood,
in BSD (POSIX locks) neterr() should return .f.
USE TEMP NEW ALIAS TEMP1 SHARED
USE TEMP NEW ALIAS TEMP2 SHARED
CLOSE
USE TEMP NEW EXCLUSIVE
?NETERR() // equal .t.
After above modification the DENY flags works in Linux because we use
BSD locks to emulate them.
The modification you made breaks only POSIX locks used for all other
synchronizations like RLOCK, FLOCK, APPEND, INDEX READ/WRITE, MEMO READ/WRITE
plus on non Linux systems also DENY flags because POSIX locks are used to
emulate them.
best regards,
Przemek
|
_______________________________________________
Harbour mailing list
[email protected]
http://lists.harbour-project.org/mailman/listinfo/harbour
