At 11:31 AM 27/01/2011, Steffen Heil wrote:

>Hi
>
>> You might also need to take into consideration that Forced Writes did not
>> work AT ALL on Linux in releases prior to v.2.0.4 - ..
>
>I am aware of that, but we are using 2.0.4, so we should be fine.
>
>Yet, what makes me headaches is, that gstat will only the the forces writes
>status, if it was ealier set using gfix. This looks like gfix is adding
>something that gstat will show besides really switching on forced writes.
>Which once again might mean, that it could be possible that the real forced
>write flag is modified without modifying the header that gstat shows,
>leading to a situation where gstat reports forced writes while it is not
>really in effect...
>
>Besides these thoughts, I am still seeking for a way to confirm, that these
>databases are in forced writes mode without modifying them in any way.
>Is there no such tool? [It need to be on command line, but if it is only a
>flag (read: bit) in the first block of the file, that information would
>suffice...)

I think this is one situation where a definitive "right answer" is likely to come only from firebird-devel - given the FW issues and the fact that you are running a Linux build. Alex Peshkov (who found the FW bug and also researched the cause and provided a solution and ALSO builds the Linux packages) doesn't watch this list, AFAIK.

If you go there, this time make sure you mention that you are running the Linux server, and put that in the Subject, too. Provide the exact build number of the Fb 2.0.x package(s) exhibiting the problem and *also* the flavour and build no. of your Lx platform.

./heLen