[Monotone-devel] Fw: address@hidden : Re: [sqlite] checking 'PRAGMA integrity_check']

[Nathaniel Smith]

So this is recorded in our own list archives, here's some info on how
to write a test that we should add..,

-- Nathaniel

-- 
  /* Tell the world that we're going to be the grim
   * reaper of innocent orphaned children.
   */
-- Linux kernel 2.4.5, main.c
> --- Begin Message ---
>
>
>
> Subject: 
>
> Re: [sqlite] checking 'PRAGMA integrity_check'
>
>
>
>
> Date: 
>
> Wed, 10 May 2006 07:24:07 -0400
>
>
>
>
> Nathaniel Smith <address@hidden> wrote:
> > 
> > So, my question is -- what is a simple way to reliably corrupt a
> > sqlite db so that "PRAGMA integrity_check" will fail?
> > 
>
> For the SQLite regression tests, we open the database
> file separately using fopen() and then write bad stuff
> into the middle of the file.
>
> Here is a recipe for generating some corruption that
> PRAGMA integrity_check will detect but which will not
> cause lasting damage to the database - the damage
> can be repaired by running REINDEX - and will not cause
> an SQLITE_CORRUPT error.
>
>    1.  Create a new table:  CREATE TABLE test1(x UNIQUE);
>        The UNIQUE argument is important because we want
>        an index for this test.
>
>    2.  Look at the SQLITE_MASTER table to find the ROOTPAGE
>        for this table.  Use fopen() to make a copy of this
>        page, taking care to remember that SQLite page numbers
>        begin with 1, not 0.
>
>    3.  Insert one small row into the test table:
>        INSERT INTO test1 VALUES(1);  Commit this change.
>
>    4.  Overwrite the rootpage of the test table with the
>        copy of that page that was saved in step 2.  This
>        effectively deletes the entry you just inserted but
>        without updating the index - so that the index and
>        the table are now out of sync.
>
> --
> D. Richard Hipp   <address@hidden>
>
>
>
>
> --- End Message ---