Hello,

unfortunately the hard disk was full on a customer site, still using
Firebird 1.5.

The corrupted database files are very small, between 1 to 3 MB.

Most of the data can still be seen in a DB grid. Only at the very end
of the tables, the EOF I/O error exception is shown.

But pumping the data with my own algorithm based on IBO can rescue
much less data than seen in the grid. The EOF error seems to be
happening much earlier than with the DB grid. Not sure why.

GFIX -v and -mend both report a bad checksum.

GBAK fails due to I/O error with EOF, even after GFIX -mend.

What's the best way to fix this?

Cheers,
Tobias