Regression Tests
regression tests
test
The regression tests are a comprehensive set of tests for the SQL
implementation in PostgreSQL. They test
standard SQL operations as well as the extended capabilities of
PostgreSQL.
Running the Tests
The regression tests can be run against an already installed and
running server, or using a temporary installation within the build
tree. Furthermore, there is a parallel
and a
sequential
mode for running the tests. The
sequential method runs each test script in turn, whereas the
parallel method starts up multiple server processes to run groups
of tests in parallel. Parallel testing gives confidence that
interprocess communication and locking are working correctly. For
historical reasons, the sequential test is usually run against an
existing installation and the parallel method against a temporary
installation, but there are no technical reasons for this.
To run the regression tests after building but before installation,
type
gmake check
in the top-level directory. (Or you can change to
src/test/regress and run the command there.)
This will first build several auxiliary files, such as
some sample user-defined trigger functions, and then run the test driver
script. At the end you should see something like
======================
All 100 tests passed.
======================
or otherwise a note about which tests failed. See below before assuming that a
failure> represents a serious problem.
Because this test method runs a temporary server, it will not work
when you are the root user (since the server will not start as root).
If you already did the build as root, you do not have to start all
over. Instead, make the regression test directory writable by
some other user, log in as that user, and restart the tests.
For example
root# chmod -R a+w src/test/regress
root# chmod -R a+w contrib/spi
root# su - joeuser
joeuser$ cd top-level build directory>
joeuser$ gmake check
(The only possible security risk
here is that other
users might be able to alter the regression test results behind
your back. Use common sense when managing user permissions.)
Alternatively, run the tests after installation.
If you have configured PostgreSQL to install
into a location where an older PostgreSQL
installation already exists, and you perform gmake check>
before installing the new version, you may find that the tests fail
because the new programs try to use the already-installed shared
libraries. (Typical symptoms are complaints about undefined symbols.)
If you wish to run the tests before overwriting the old installation,
you'll need to build with configure --disable-rpath>.
It is not recommended that you use this option for the final installation,
however.
The parallel regression test starts quite a few processes under your
user ID. Presently, the maximum concurrency is twenty parallel test
scripts, which means sixty processes: there's a server process, a
psql>, and usually a shell parent process for the
psql> for each test script.
So if your system enforces a per-user limit on the number of processes,
make sure this limit is at least seventy-five or so, else you may get
random-seeming failures in the parallel test. If you are not in
a position to raise the limit, you can cut down the degree of parallelism
by setting the MAX_CONNECTIONS> parameter. For example,
gmake MAX_CONNECTIONS=10 check
runs no more than ten tests concurrently.
On some systems, the default Bourne-compatible shell
(/bin/sh) gets confused when it has to manage
too many child processes in parallel. This may cause the parallel
test run to lock up or fail. In such cases, specify a different
Bourne-compatible shell on the command line, for example:
gmake SHELL=/bin/ksh check
If no non-broken shell is available, you may be able to work around the
problem by limiting the number of connections, as shown above.
To run the tests after installation)]]>,
initialize a data area and start the
server, , ]]> then type
gmake installcheck
or for a parallel test
gmake installcheck-parallel
The tests will expect to contact the server at the local host and the
default port number, unless directed otherwise by PGHOST and
PGPORT environment variables.
The source distribution also contains regression tests for the optional
procedural languages and for some of the contrib> modules.
At present, these tests can be used only against an already-installed
server. To run the tests for all procedural languages that have been
built and installed, change to the src/pl> directory of the
build tree and type
gmake installcheck
You can also do this in any of the subdirectories of src/pl>
to run tests for just one procedural language. To run the tests for all
contrib> modules that have them, change to the
contrib> directory of the build tree and type
gmake installcheck
The contrib> modules must have been built and installed first.
You can also do this in a subdirectory of contrib> to run
the tests for just one module.
Test Evaluation
Some properly installed and fully functional
PostgreSQL installations can
fail
some of these regression tests due to
platform-specific artifacts such as varying floating-point representation
and message wording. The tests are currently evaluated using a simple
diff comparison against the outputs
generated on a reference system, so the results are sensitive to
small system differences. When a test is reported as
failed
, always examine the differences between
expected and actual results; you may well find that the
differences are not significant. Nonetheless, we still strive to
maintain accurate reference files across all supported platforms,
so it can be expected that all tests pass.
The actual outputs of the regression tests are in files in the
src/test/regress/results directory. The test
script uses diff to compare each output
file against the reference outputs stored in the
src/test/regress/expected directory. Any
differences are saved for your inspection in
src/test/regress/regression.diffs. (Or you
can run diff yourself, if you prefer.)
If for some reason a particular platform generates a failure>
for a given test, but inspection of the output convinces you that
the result is valid, you can add a new comparison file to silence
the failure report in future test runs. See
for details.
Error message differences
Some of the regression tests involve intentional invalid input
values. Error messages can come from either the
PostgreSQL code or from the host
platform system routines. In the latter case, the messages may
vary between platforms, but should reflect similar
information. These differences in messages will result in a
failed
regression test that can be validated by
inspection.
Locale differences
If you run the tests against an already-installed server that was
initialized with a collation-order locale other than C, then
there may be differences due to sort order and follow-up
failures. The regression test suite is set up to handle this
problem by providing alternative result files that together are
known to handle a large number of locales.
Date and time differences
Most of the date and time results are dependent on the time zone
environment. The reference files are generated for time zone
PST8PDT (Berkeley, California), and there will be
apparent failures if the tests are not run with that time zone setting.
The regression test driver sets environment variable
PGTZ to PST8PDT, which normally
ensures proper results.
Floating-point differences
Some of the tests involve computing 64-bit floating-point numbers (double
precision) from table columns. Differences in
results involving mathematical functions of double
precision columns have been observed. The float8> and
geometry> tests are particularly prone to small differences
across platforms, or even with different compiler optimization options.
Human eyeball comparison is needed to determine the real
significance of these differences which are usually 10 places to
the right of the decimal point.
Some systems display minus zero as -0>, while others
just show 0>.
Some systems signal errors from pow() and
exp() differently from the mechanism
expected by the current PostgreSQL
code.
Row ordering differences
You might see differences in which the same rows are output in a
different order than what appears in the expected file. In most cases
this is not, strictly speaking, a bug. Most of the regression test
scripts are not so pedantic as to use an ORDER BY> for every single
SELECT>, and so their result row orderings are not well-defined
according to the letter of the SQL specification. In practice, since we are
looking at the same queries being executed on the same data by the same
software, we usually get the same result ordering on all platforms, and
so the lack of ORDER BY> isn't a problem. Some queries do exhibit
cross-platform ordering differences, however. When testing against an
already-installed server, ordering differences can also be caused by
non-C locale settings or non-default parameter settings, such as custom values
of work_mem> or the planner cost parameters.
Therefore, if you see an ordering difference, it's not something to
worry about, unless the query does have an ORDER BY> that your
result is violating. But please report it anyway, so that we can add an
ORDER BY> to that particular query and thereby eliminate the bogus
failure
in future releases.
You might wonder why we don't order all the regression test queries explicitly
to get rid of this issue once and for all. The reason is that that would
make the regression tests less useful, not more, since they'd tend
to exercise query plan types that produce ordered results to the
exclusion of those that don't.
Insufficient stack depth
If the errors test results in a server crash
at the select infinite_recurse()> command, it means that
the platform's limit on process stack size is smaller than the
]]>
max_stack_depth]]>
parameter indicates. This
can be fixed by running the server under a higher stack
size limit (4MB is recommended with the default value of
max_stack_depth>). If you are unable to do that, an
alternative is to reduce the value of max_stack_depth>.
The random
test
The random test script is intended to produce
random results. In rare cases, this causes the random regression
test to fail. Typing
diff results/random.out expected/random.out
should produce only one or a few lines of differences. You need
not worry unless the random test fails repeatedly.
Variant Comparison Files
Since some of the tests inherently produce environment-dependent
results, we have provided ways to specify alternative expected>
result files. Each regression test can have several comparison files
showing possible results on different platforms. There are two
independent mechanisms for determining which comparison file is used
for each test.
The first mechanism allows comparison files to be selected for
specific platforms. There is a mapping file,
src/test/regress/resultmap, that defines
which comparison file to use for each platform.
To eliminate bogus test failures
for a particular platform,
you first choose or make a variant result file, and then add a line to the
resultmap file.
Each line in the mapping file is of the form
testname/platformpattern=comparisonfilename
The test name is just the name of the particular regression test
module. The platform pattern is a pattern in the style of the Unix
tool expr> (that is, a regular expression with an implicit
^ anchor
at the start). It is matched against the platform name as printed
by config.guess followed by
:gcc or :cc, depending on
whether you use the GNU compiler or the system's native compiler
(on systems where there is a difference). The comparison file
name is the base name of the substitute result comparison file.
For example: some systems interpret very small floating-point values
as zero, rather than reporting an underflow error. This causes a
few differences in the float8> regression test.
Therefore, we provide a variant comparison file,
float8-small-is-zero.out, which includes
the results to be expected on these systems. To silence the bogus
failure
message on OpenBSD
platforms, resultmap includes
float8/i.86-.*-openbsd=float8-small-is-zero
which will trigger on any machine for which the output of
config.guess matches i.86-.*-openbsd.
Other lines
in resultmap> select the variant comparison file for other
platforms where it's appropriate.
The second selection mechanism for variant comparison files is
much more automatic: it simply uses the best match> among
several supplied comparison files. The regression test driver
script considers both the standard comparison file for a test,
testname>.out>, and variant files named
testname>_digit>.out>
(where the digit> is any single digit
0>-9>). If any such file is an exact match,
the test is considered to pass; otherwise, the one that generates
the shortest diff is used to create the failure report. (If
resultmap includes an entry for the particular
test, then the base testname> is the substitute
name given in resultmap.)
For example, for the char test, the comparison file
char.out contains results that are expected
in the C> and POSIX> locales, while
the file char_1.out contains results sorted as
they appear in many other locales.
The best-match mechanism was devised to cope with locale-dependent
results, but it can be used in any situation where the test results
cannot be predicted easily from the platform name alone. A limitation of
this mechanism is that the test driver cannot tell which variant is
actually correct> for the current environment; it will just pick
the variant that seems to work best. Therefore it is safest to use this
mechanism only for variant results that you are willing to consider
equally valid in all contexts.