8 1. Change to always load suites, then resolve command line arguments?
10 Currently we expect each input argument to be a path on disk; we do a
11 recursive search to find the test suite for each item, but then we only do a
12 local search based at the input path to find tests. Additionally, for any path
13 that matches a file on disk we explicitly construct a test instance (bypassing
14 the formats on discovery implementation).
16 This has a couple problems:
18 * The test format doesn't have control over the test instances that result
21 * It isn't possible to specify virtual tests as inputs. For example, it is not
22 possible to specify an individual subtest to run with the googletest format.
24 * The test format doesn't have full control over the discovery of tests in
27 Instead, we should move to a model whereby first all of the input specifiers
28 are resolved to test suites, and then the resolution of the input specifier is
29 delegated to each test suite. This could take a couple forms:
31 * We could resolve to test suites, then fully load each test suite, then have
32 a fixed process to map input specifiers to tests in the test suite
33 (presumably based on path-in-suite derivations). This has the benefit of
34 being consistent across all test formats, but the downside of requiring
35 loading the entire test suite.
37 * We could delegate all of the resolution of specifiers to the test
38 suite. This would allow formats that anticipate large test suites to manage
39 their own resolution for better performance. We could provide a default
40 resolution strategy that was similar to what we do now (start at subpaths
41 for directories, but allow the test format control over what happens for
44 2. Consider move to identifying all tests by path-to-test-suite and then path to
45 subtest, and don't use test suite names.
47 Currently the test suite name is presented as part of test names, but it has
48 no other useful function, and it is something that has to be skipped over to
49 cut-and-paste a name to subsequently use to rerun a test. If we just
50 represented each test suite by the path to its suite, then it would allow more
51 easy cut-and-paste of the test output lines. This has the downside that the
52 lines might get rather long.
54 3. Allow 'lit' driver to cooperate with test formats and suites to add options
55 (or at least sanitize accepted params).
57 We have started to use the --params method more and more extensively, and it is
58 cumbersome and error prone. Additionally, there are currently various options
59 ``lit`` honors that should more correctly be specified as belonging to the
62 It would be really nice if we could allow test formats and test suites to add
63 their own options to be parsed. The difficulty here, of course, is that we
64 don't know what test formats or test suites are in use until we have parsed the
65 input specifiers. For test formats we could ostensibly require all the possible
66 formats to be registered in order to have options, but for test suites we would
67 certainly have to load the suite before we can query it for what options it
70 That leaves us with the following options:
72 * Currently we could almost get away with parsing the input specifiers without
73 having done option parsing first (the exception is ``--config-prefix``) but
74 that isn't a very extensible design.
76 * We could make a distinction in the command line syntax for test format and
77 test suite options. For example, we could require something like::
79 lit -j 1 -sv input-specifier -- --some-format-option
81 which would be relatively easy to implement with optparser (I think).
83 * We could allow fully interspersed arguments by first extracting the options
84 lit knows about and parsing them, then dispatching the remainder to the
85 formats. This seems the most convenient for users, who are unlikely to care
86 about (or even be aware of) the distinction between the generic lit
87 infrastructure and format or suite specific options.
89 4. Eliminate duplicate execution models for ShTest tests.
91 Currently, the ShTest format uses tests written with shell-script like syntax,
92 and executes them in one of two ways. The first way is by converting them into
93 a bash script and literally executing externally them using bash. The second
94 way is through the use of an internal shell parser and shell execution code
95 (built on the subprocess module). The external execution mode is used on most
96 Unix systems that have bash, the internal execution mode is used on Windows.
98 Having two ways to do the same thing is error prone and leads to unnecessary
99 complexity in the testing environment. Additionally, because the mode that
100 converts scripts to bash doesn't try and validate the syntax, it is possible
101 to write tests that use bash shell features unsupported by the internal
102 shell. Such tests won't work on Windows but this may not be obvious to the
103 developer writing the test.
105 Another limitation is that when executing the scripts externally, the ShTest
106 format has no idea which commands fail, or what output comes from which
107 commands, so this limits how convenient the output of ShTest failures can be
108 and limits other features (for example, knowing what temporary files were
111 We should eliminate having two ways of executing the same tests to reduce
112 platform differences and make it easier to develop new features in the ShTest
113 module. This is currently blocked on:
115 * The external execution mode is faster in some situations, because it avoids
116 being bottlenecked on the GIL. This can hopefully be obviated simply by
117 using --use-processes.
119 * Some tests in LLVM/Clang are explicitly disabled with the internal shell
120 (because they use features specific to bash). We would need to rewrite these
121 tests, or add additional features to the internal shell handling to allow
124 5. Consider changing core to support setup vs. execute distinction.
126 Many of the existing test formats are cleanly divided into two phases, once
127 parses the test format and extracts XFAIL and REQUIRES information, etc., and
128 the other code actually executes the test.
130 We could make this distinction part of the core infrastructure and that would
131 enable a couple things:
133 * The REQUIREs handling could be lifted to the core, which is nice.
135 * This would provide a clear place to insert subtest support, because the
136 setup phase could be responsible for providing subtests back to the
137 core. That would provide part of the infrastructure to parallelize them, for
138 example, and would probably interact well with other possible features like
141 * This affords a clean implementation of --no-execute.
143 * One possible downside could be for test formats that cannot determine their
144 subtests without having executed the test. Supporting such formats would
145 either force the test to actually be executed in the setup stage (which
146 might be ok, as long as the API was explicitly phrased to support that), or
147 would mean we are forced into supporting subtests as return values from the
150 Any format can just keep all of its code in execute, presumably, so the only
151 cost of implementing this is its impact on the API and futures changes.
157 * Move temp directory name into local test config.
159 * Add --show-unsupported, don't show by default?
161 * Support valgrind in all configs, and LLVM style valgrind.
163 * Support a timeout / ulimit.
165 * Create an explicit test suite object (instead of using the top-level
166 TestingConfig object).