Regenerate badly-obsolete README file.

This commit is contained in:
Tom Lane 2005-07-24 17:18:52 +00:00
parent 4c80071b81
commit d675226e20

View File

@ -1,280 +1,278 @@
Regression Tests Regression Tests
The regression tests are a comprehensive set of tests for the SQL The regression tests are a comprehensive set of tests for the SQL
implementation in PostgreSQL. They test standard SQL operations as well as the implementation in PostgreSQL. They test standard SQL operations as
extended capabilities of PostgreSQL. From PostgreSQL 6.1 onward, the regression well as the extended capabilities of PostgreSQL.
tests are current for every official release. _________________________________________________________________
------------------------------------------------------------------------------- Running the Tests
Running the Tests The regression tests can be run against an already installed and
running server, or using a temporary installation within the build
The regression test can be run against an already installed and running server, tree. Furthermore, there is a "parallel" and a "sequential" mode for
or using a temporary installation within the build tree. Furthermore, there is running the tests. The sequential method runs each test script in
a "parallel" and a "sequential" mode for running the tests. The sequential turn, whereas the parallel method starts up multiple server processes
method runs each test script in turn, whereas the parallel method starts up to run groups of tests in parallel. Parallel testing gives confidence
multiple server processes to run groups of tests in parallel. Parallel testing that interprocess communication and locking are working correctly. For
gives confidence that interprocess communication and locking are working historical reasons, the sequential test is usually run against an
correctly. For historical reasons, the sequential test is usually run against existing installation and the parallel method against a temporary
an existing installation and the parallel method against a temporary installation, but there are no technical reasons for this.
installation, but there are no technical reasons for this.
To run the regression tests after building but before installation,
To run the regression tests after building but before installation, type type
gmake check
gmake check
in the top-level directory. (Or you can change to "src/test/regress"
in the top-level directory. (Or you can change to "src/test/regress" and run and run the command there.) This will first build several auxiliary
the command there.) This will first build several auxiliary files, such as some files, such as some sample user-defined trigger functions, and then
sample user-defined trigger functions, and then run the test driver script. At run the test driver script. At the end you should see something like
the end you should see something like ======================
All 98 tests passed.
====================== ======================
All 93 tests passed.
====================== or otherwise a note about which tests failed. See the section called
Test Evaluation below before assuming that a "failure" represents a
or otherwise a note about which tests failed. See the Section called Test serious problem.
Evaluation below for more.
Because this test method runs a temporary server, it will not work
Because this test method runs a temporary server, it will not work when you are when you are the root user (since the server will not start as root).
the root user (since the server will not start as root). If you already did the If you already did the build as root, you do not have to start all
build as root, you do not have to start all over. Instead, make the regression over. Instead, make the regression test directory writable by some
test directory writable by some other user, log in as that user, and restart other user, log in as that user, and restart the tests. For example
the tests. For example root# chmod -R a+w src/test/regress
root# chmod -R a+w contrib/spi
root# chmod -R a+w src/test/regress root# su - joeuser
root# chmod -R a+w contrib/spi joeuser$ cd top-level build directory
root# su - joeuser joeuser$ gmake check
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
(The only possible "security risk" here is that other users might be able to sense when managing user permissions.)
alter the regression test results behind your back. Use common sense when
managing user permissions.) Alternatively, run the tests after installation.
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
The parallel regression test starts quite a few processes under your user ID. check before installing the new version, you may find that the tests
Presently, the maximum concurrency is twenty parallel test scripts, which means fail because the new programs try to use the already-installed shared
sixty processes: there's a server process, a psql, and usually a shell parent libraries. (Typical symptoms are complaints about undefined symbols.)
process for the psql for each test script. So if your system enforces a per- If you wish to run the tests before overwriting the old installation,
user limit on the number of processes, make sure this limit is at least you'll need to build with configure --disable-rpath. It is not
seventy-five or so, else you may get random-seeming failures in the parallel recommended that you use this option for the final installation,
test. If you are not in a position to raise the limit, you can cut down the however.
degree of parallelism by setting the MAX_CONNECTIONS parameter. For example,
The parallel regression test starts quite a few processes under your
gmake MAX_CONNECTIONS=10 check user ID. Presently, the maximum concurrency is twenty parallel test
scripts, which means sixty processes: there's a server process, a
runs no more than ten tests concurrently. 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
On some systems, the default Bourne-compatible shell ("/bin/sh") gets confused processes, make sure this limit is at least seventy-five or so, else
when it has to manage too many child processes in parallel. This may cause the you may get random-seeming failures in the parallel test. If you are
parallel test run to lock up or fail. In such cases, specify a different not in a position to raise the limit, you can cut down the degree of
Bourne-compatible shell on the command line, for example: parallelism by setting the MAX_CONNECTIONS parameter. For example,
gmake MAX_CONNECTIONS=10 check
gmake SHELL=/bin/ksh check
runs no more than ten tests concurrently.
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. On some systems, the default Bourne-compatible shell ("/bin/sh") gets
confused when it has to manage too many child processes in parallel.
To run the tests after installation, initialize a data area and start the This may cause the parallel test run to lock up or fail. In such
server, then type cases, specify a different Bourne-compatible shell on the command
line, for example:
gmake installcheck gmake SHELL=/bin/ksh check
The tests will expect to contact the server at the local host and the default If no non-broken shell is available, you may be able to work around
port number, unless directed otherwise by PGHOST and PGPORT environment the problem by limiting the number of connections, as shown above.
variables.
To run the tests after installation, initialize a data area and start
------------------------------------------------------------------------------- the server, then type
gmake installcheck
Test Evaluation
or for a parallel test
Some properly installed and fully functional PostgreSQL installations can gmake installcheck-parallel
"fail" some of these regression tests due to platform-specific artifacts such
as varying floating-point representation and time zone support. The tests are The tests will expect to contact the server at the local host and the
currently evaluated using a simple "diff" comparison against the outputs default port number, unless directed otherwise by PGHOST and PGPORT
generated on a reference system, so the results are sensitive to small system environment variables.
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 Test Evaluation
accurate reference files across all supported platforms, so it can be expected
that all tests pass. Some properly installed and fully functional PostgreSQL installations
can "fail" some of these regression tests due to platform-specific
The actual outputs of the regression tests are in files in the "src/test/ artifacts such as varying floating-point representation and message
regress/results" directory. The test script uses "diff" to compare each output wording. The tests are currently evaluated using a simple "diff"
file against the reference outputs stored in the "src/test/regress/expected" comparison against the outputs generated on a reference system, so the
directory. Any differences are saved for your inspection in "src/test/regress/ results are sensitive to small system differences. When a test is
regression.diffs". (Or you can run "diff" yourself, if you prefer.) 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
Error message differences that all tests pass.
Some of the regression tests involve intentional invalid input values. Error The actual outputs of the regression tests are in files in the
messages can come from either the PostgreSQL code or from the host platform "src/test/regress/results" directory. The test script uses "diff" to
system routines. In the latter case, the messages may vary between platforms, compare each output file against the reference outputs stored in the
but should reflect similar information. These differences in messages will "src/test/regress/expected" directory. Any differences are saved for
result in a "failed" regression test that can be validated by inspection. 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
Locale differences 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
If you run the tests against an already-installed server that was initialized report in future test runs. See the section called Variant Comparison
with a collation-order locale other than C, then there may be differences due Files for details.
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. For example, for the char test, the Error message differences
expected file "char.out" handles the C and POSIX locales, and the file
"char_1.out" handles many other locales. The regression test driver will Some of the regression tests involve intentional invalid input values.
automatically pick the best file to match against when checking for success and Error messages can come from either the PostgreSQL code or from the
for computing failure differences. (This means that the regression tests cannot host platform system routines. In the latter case, the messages may
detect whether the results are appropriate for the configured locale. The tests vary between platforms, but should reflect similar information. These
will simply pick the one result file that works best.) differences in messages will result in a "failed" regression test that
can be validated by inspection.
If for some reason the existing expected files do not cover some locale, you _________________________________________________________________
can add a new file. The naming scheme is testname_digit.out. The actual digit
is not significant. Remember that the regression test driver will consider all Locale differences
such files to be equally valid test results. If the test results are platform-
specific, the technique described in the Section called Platform-specific If you run the tests against an already-installed server that was
comparison files should be used instead. 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
Date and time differences number of locales.
_________________________________________________________________
A few of the queries in the "horology" test will fail if you run the test on
the day of a daylight-saving time changeover, or the day after one. These Date and time differences
queries expect that the intervals between midnight yesterday, midnight today
and midnight tomorrow are exactly twenty-four hours --- which is wrong if Most of the date and time results are dependent on the time zone
daylight-saving time went into or out of effect meanwhile. environment. The reference files are generated for time zone PST8PDT
(Berkeley, California), and there will be apparent failures if the
Note: Because USA daylight-saving time rules are used, this problem tests are not run with that time zone setting. The regression test
always occurs on the first Sunday of April, the last Sunday of driver sets environment variable PGTZ to PST8PDT, which normally
October, and their following Mondays, regardless of when daylight- ensures proper results.
saving time is in effect where you live. Also note that the problem _________________________________________________________________
appears or disappears at midnight Pacific time (UTC-7 or UTC-8), not
midnight your local time. Thus the failure may appear late on Floating-point differences
Saturday or persist through much of Tuesday, depending on where you
live. Some of the tests involve computing 64-bit floating-point numbers
(double precision) from table columns. Differences in results
Most of the date and time results are dependent on the time zone environment. involving mathematical functions of double precision columns have been
The reference files are generated for time zone PST8PDT (Berkeley, California), observed. The float8 and geometry tests are particularly prone to
and there will be apparent failures if the tests are not run with that time small differences across platforms, or even with different compiler
zone setting. The regression test driver sets environment variable PGTZ to optimization options. Human eyeball comparison is needed to determine
PST8PDT, which normally ensures proper results. However, your operating system the real significance of these differences which are usually 10 places
must provide support for the PST8PDT time zone, or the time zone-dependent to the right of the decimal point.
tests will fail. To verify that your machine does have this support, type the
following: Some systems display minus zero as -0, while others just show 0.
env TZ=PST8PDT date Some systems signal errors from pow() and exp() differently from the
mechanism expected by the current PostgreSQL code.
The command above should have returned the current system time in the PST8PDT _________________________________________________________________
time zone. If the PST8PDT time zone is not available, then your system may have
returned the time in UTC. If the PST8PDT time zone is missing, you can set the Row ordering differences
time zone rules explicitly:
You might see differences in which the same rows are output in a
PGTZ='PST8PDT7,M04.01.0,M10.05.03'; export PGTZ different order than what appears in the expected file. In most cases
this is not, strictly speaking, a bug. Most of the regression test
There appear to be some systems that do not accept the recommended syntax for scripts are not so pedantic as to use an ORDER BY for every single
explicitly setting the local time zone rules; you may need to use a different SELECT, and so their result row orderings are not well-defined
PGTZ setting on such machines. 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
Some systems using older time-zone libraries fail to apply daylight-saving the same software, we usually get the same result ordering on all
corrections to dates before 1970, causing pre-1970 PDT times to be displayed in platforms, and so the lack of ORDER BY isn't a problem. Some queries
PST instead. This will result in localized differences in the test results. 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.
Floating-point differences
Therefore, if you see an ordering difference, it's not something to
Some of the tests involve computing 64-bit floating-point numbers (double worry about, unless the query does have an ORDER BY that your result
precision) from table columns. Differences in results involving mathematical is violating. But please report it anyway, so that we can add an ORDER
functions of double precision columns have been observed. The float8 and BY to that particular query and thereby eliminate the bogus "failure"
geometry tests are particularly prone to small differences across platforms, or in future releases.
even with different compiler optimization options. Human eyeball comparison is
needed to determine the real significance of these differences which are You might wonder why we don't order all the regression test queries
usually 10 places to the right of the decimal point. 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
Some systems display minus zero as -0, while others just show 0. they'd tend to exercise query plan types that produce ordered results
to the exclusion of those that don't.
Some systems signal errors from pow() and exp() differently from the mechanism _________________________________________________________________
expected by the current PostgreSQL code.
The "random" test
-------------------------------------------------------------------------------
The random test script is intended to produce random results. In rare
Row ordering differences cases, this causes the random regression test to fail. Typing
diff results/random.out expected/random.out
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, should produce only one or a few lines of differences. You need not
strictly speaking, a bug. Most of the regression test scripts are not so worry unless the random test fails repeatedly.
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 Variant Comparison Files
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 Since some of the tests inherently produce environment-dependent
queries do exhibit cross-platform ordering differences, however. (Ordering results, we have provided ways to specify alternative "expected"
differences can also be triggered by non-C locale settings.) result files. Each regression test can have several comparison files
showing possible results on different platforms. There are two
Therefore, if you see an ordering difference, it's not something to worry independent mechanisms for determining which comparison file is used
about, unless the query does have an ORDER BY that your result is violating. for each test.
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. The first mechanism allows comparison files to be selected for
specific platforms. There is a mapping file,
You might wonder why we don't order all the regression test queries explicitly "src/test/regress/resultmap", that defines which comparison file to
to get rid of this issue once and for all. The reason is that that would make use for each platform. To eliminate bogus test "failures" for a
the regression tests less useful, not more, since they'd tend to exercise query particular platform, you first choose or make a variant result file,
plan types that produce ordered results to the exclusion of those that don't. and then add a line to the "resultmap" file.
------------------------------------------------------------------------------- Each line in the mapping file is of the form
testname/platformpattern=comparisonfilename
The "random" test
The test name is just the name of the particular regression test
There is at least one case in the random test script that is intended to module. The platform pattern is a pattern in the style of the Unix
produce random results. This causes random to fail the regression test once in tool "expr" (that is, a regular expression with an implicit ^ anchor
a while (perhaps once in every five to ten trials). Typing 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
diff results/random.out expected/random.out 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
should produce only one or a few lines of differences. You need not worry the substitute result comparison file.
unless the random test always fails in repeated attempts. (On the other hand,
if the random test is *never* reported to fail even in many trials of the For example: some systems interpret very small floating-point values
regression tests, you probably *should* worry.) 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
Platform-specific comparison files "failure" message on OpenBSD platforms, "resultmap" includes
float8/i.86-.*-openbsd=float8-small-is-zero
Since some of the tests inherently produce platform-specific results, we have
provided a way to supply platform-specific result comparison files. Frequently, which will trigger on any machine for which the output of
the same variation applies to multiple platforms; rather than supplying a "config.guess" matches i.86-.*-openbsd. Other lines in "resultmap"
separate comparison file for every platform, there is a mapping file that select the variant comparison file for other platforms where it's
defines which comparison file to use. So, to eliminate bogus test "failures" appropriate.
for a particular platform, you must choose or make a variant result file, and
then add a line to the mapping file, which is "src/test/regress/resultmap". The second selection mechanism for variant comparison files is much
more automatic: it simply uses the "best match" among several supplied
Each line in the mapping file is of the form comparison files. The regression test driver script considers both the
standard comparison file for a test, testname.out, and variant files
testname/platformpattern=comparisonfilename 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;
The test name is just the name of the particular regression test module. The otherwise, the one that generates the shortest diff is used to create
platform pattern is a pattern in the style of the Unix tool "expr" (that is, a the failure report. (If "resultmap" includes an entry for the
regular expression with an implicit ^ anchor at the start). It is matched particular test, then the base "testname" is the substitute name given
against the platform name as printed by "config.guess" followed by :gcc or :cc, in "resultmap".)
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 name For example, for the char test, the comparison file "char.out"
of the substitute result comparison file. 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
For example: some systems using older time zone libraries fail to apply other locales.
daylight-saving corrections to dates before 1970, causing pre-1970 PDT times to
be displayed in PST instead. This causes a few differences in the "horology" The best-match mechanism was devised to cope with locale-dependent
regression test. Therefore, we provide a variant comparison file, "horology-no- results, but it can be used in any situation where the test results
DST-before-1970.out", which includes the results to be expected on these cannot be predicted easily from the platform name alone. A limitation
systems. To silence the bogus "failure" message on HPUX platforms, "resultmap" of this mechanism is that the test driver cannot tell which variant is
includes actually "correct" for the current environment; it will just pick the
variant that seems to work best. Therefore it is safest to use this
horology/.*-hpux=horology-no-DST-before-1970 mechanism only for variant results that you are willing to consider
equally valid in all contexts.
which will trigger on any machine for which the output of "config.guess"
includes -hpux. Other lines in "resultmap" select the variant comparison file
for other platforms where it's appropriate.