Jean Francois Ortolo wrote:
>
> For my own, I did the second e.g. physical check of the surface,
> then reported the results in one of my recent posts, that is:
> - no bad block
> - there were disk changes being made.
Oops, I missed that, sorry!
> Yes, I keep wondering how this, and only this file, may have been
> corrupted in my filesystem ( /dev/hda5 ). This happened while I tried
> to run mysql from an xterm. Just before that, mysql was working fine
> without any problem.
>
> I don't believe either a program, or a library, can write in the
> library physical file. This is not possible, as much as my knowledge
> comes.
root can do (nearly) anything. This is not Windows, in Unix there is no
method to *force* other processes not to touch a file. You can ask them
via fcntl() or flags in open(), but the other process can just ignore
this.
A change to the library file doesn't disturb a running mysqld: When
mysqld is started all needed shared libraries are loaded into memory (if
they aren't already there). A library will only be unloaded from memory
when the last process using it is terminated. In the meantime the
library file might have been altered by another process or maybe a
renegade harddisk head. This will not be noted because unloading the
library from memory means the memory image will just be discarded. There
will be no comparison between the memory image and the file, also the
memory image will not be written back onto the harddisk, it is just
discarded.
If you kept the defective library file in a backup place you can compare
the defective and the correct file and might get a clue about what
happened.
Best regards,
Martin Stricker
--
Homepage: http://www.martin-stricker.de/
Registered Linux user #210635: http://counter.li.org/
_______________________________________________
Seawolf-list mailing list
[EMAIL PROTECTED]
https://listman.redhat.com/mailman/listinfo/seawolf-list
