You can only unset it if the TDB was originally opened O_RDWR.

Also, cleaned up error handling in tdb_allrecord_lock() so we only get
one log message on a r/o database.

This is more accurate than returning TDB_ERR_EINVAL.

It's more consistent with what tdb1 does, and slightly more encapsulated.

Since we've deprecated tdb_error() function (and it didn't work right
from inside the logging function anyway, since we didn't set
tdb->ecode yet) we need to hand it to the log function.

Make it the same offset as TDB1.  This isn't strictly necessary, but
it would allow for total unification later, since TDB1 and TDB2's
transaction code is otherwise completely compatible.

If the caller actually specifies TDB_CONVERT, fail if the TDB does not
need conversion.  It's unusual for the user to specify TDB_CONVERT
(it's auto-detected) but better to be strict.

The tdb_oob() function can fail due to errors, as well as because the length
asked for is greater than the size of the file.  Clean that up:

(1) If probe is true, only fail if there's an error, not if the length is
    too great.
(2) Exit tdb_open() if it tdb_oob() probe fails; this helps cut down
    test time for failtest.
(3) Don't set probe to true in tdb_direct() fail; a minor issue, but it means
    we log failure.

It can be 64 bits, for huge databases.

As noted, failtest was taking a long time, because a failure injected here
was not detected.

We use off_t in the tdb2 interface (for tdb_attribute_flock); we need
to make sure that all callers agree on the size.

This also causes a problem in the tests: it's not enough to include
config.h first, we need the _FILE_OFFSET_BITS define from private.h.
Otherwise, we can disagree about the definitions of F_SETLK, F_SETLKW
and off_t, causing strange problems.

Autobuild-User: Volker Lendecke <vlendec@samba.org>
Autobuild-Date: Wed Aug 10 21:12:59 CEST 2011 on sn-devel-104

(Imported from SAMBA commit f7b820c3590ac78dd0dea67ac882f53ae6f550d9)

(Imported from SAMBA commit fed977514019e130322fb21ad5726ee997965898)

(Imported from SAMBA commit 35550ba641d251a2dc3762b08eda92d0f7ce6bed)

(Imported from SAMBA commit eac2f44cde39d16b827b7e2d360f7a7fb1d940e5)

(Imported from SAMBA commit 06a41594554a1a6881dcbf9d051a822f8f955b56)

(Imported from SAMBA commit f589af19821e12aba6145492fd4ded58f07131b8)

(Imported from SAMBA commit b32f086d643eb39c3806e30b19a028e6a59f0054)

(Imported from SAMBA commit ee91004f4bddcda873f8ccde1db16609a66a493b)

(Imported from SAMBA commit 26795b3dd1a90c61424cd8f1761f98bc60b3ddc4)

(Imported from SAMBA commit 73414d0ee7bbe371b0f9b275b44ce5854a10b9f2)

(Imported from SAMBA commit 3f67063dfab1a7add95da5157206ccebb2f75df8)

(Imported from SAMBA commit 9c32c86783bf6a3e126e6e83f51ccdf0ecdd132d)

(Imported from SAMBA commit 1fa4236c40a727d7ae648e12d14d1f0f9fcf79b5)

(Imported from SAMBA commit 4a3913fbb9399685785efa430765e0b5836ded60)

(Imported from SAMBA commit a2d1d0b51485f2d383d1d69fa3bd6b79b9e8d5aa)

(Imported from SAMBA commit 68d0843437b05dc8a36091e4a307bfd912cbc09f)

(Imported from SAMBA commit c6ae297d613fc22b92c34d26c61ec1715058b484)

(Imported from SAMBA commit 066d36a1a635e1115f62c452c49a9830d484c03b)

(Imported from SAMBA commit cb5c6f441f394f91bedf641aa76841bdb833e440)

(Imported from SAMBA commit 54282e9f4eda083e70c7e56dda2bf425209add6d)

Rather than timeout, explicitly disable tests_pass_valgrind and
tests_compile_coverage.  This is important now we run under valgrind
the first time we run the tests, thus would always time out and not
run most of the tests at all!

"make check" takes 7m57s, old "make fastcheck" takes 3m28s, and this takes
only 1m57s, but much more is run (as shown by the larger score total).

Putting the reduced config.h in the current directory means that it's
actually being picked up by other tests, such as the string checks.
So move it to a sub-directory where we need an explicit -I.

We also fix the dependencies, so that "--target
tests_pass_without_features" works.

Tests can be slow to run, and this way we can monitor progress.

We eliminate dependencies as we recurse, but if a single _info file
lists a dependency twice, we add it to the list twice and this skip
over the middle ones.

Rather than a separate cov_compiled member, we can add to the
compiled[] array, and we reduce duplication significantly.

Rather than mug the old ->compiled version when we reduce features,
keep both in the structure.  This makes it clear that we are using the
right version (we weren't in all cases, in particular we weren't
recompiling the test helpers correctly.

Unless it's not available or suppressed, we might as well run it under
valgrind the first time, which speeds things a little.

We save the valgrind output file for tests_pass_valgrind's use later.

We'll need this in a moment.

There's no reason to hand in the name of the compiled _info file,
since we don't compile it up in safe mode.