Annotation of wikisrc/kyua/import.mdwn, revision 1.2

1.1       jmmv        1: [[!meta title="Kyua: The way into NetBSD"]]
                      2: [[!toc levels=2]]
                      4: **Project owner: [Julio Merino](**  
                      5: **Status: draft as of 2012-08-31; not even sent for review yet.**
                      7: The import of Kyua into NetBSD to replace the deprecated ATF tools is
                      8: planned to happen in NetBSD 7.0.  The ATF libraries will remain in place,
                      9: and as such no changes will happen to the tests that already live in
                     10: `src/tests`.
                     12: The transition from ATF to Kyua includes steps and tools to offer backwards
                     13: compatibility with users that may rely on the ATF tools shipped since
                     14: NetBSD 5.0.  These backwards compatibility tools cover the most common use
                     15: cases but might not be perfect initially.  Despite these provisions, this
                     16: plan should be executed well in advance the creation of the 7.0 branch to
                     17: ensure there is enough time in NetBSD-current to flesh out any major
                     18: problems.
                     20: Before moving on, please read the
1.2     ! jmmv       21: [[Kyua: An introduction for NetBSD users|/kyua]] page.  You should be
1.1       jmmv       22: familiar with the structure of Kyua and its major components to be able to
                     23: review this plan.
                     25: # Proposed changes
                     27: The final user visible changes that this project will bring to NetBSD are
                     28: as follows.  Because these change the availability of tools that have
                     29: already been shipped since 5.0, the modifications are staged to happen
                     30: across two major releases:
                     32: ## For NetBSD 7.0
                     34: * Addition of the new `kyua` command-line tool.  This provides the runtime
                     35:   engine for the test programs (`atf-run`'s functionality) and the ability
                     36:   to generate reports of the results in plain text form and HTML form
                     37:   (`atf-report`'s functionality).  This tool is able to execute all the
                     38:   existing test programs without modifications.
                     40: * Replacement of the `Atffile`s files in `/usr/tests` with `Kyuafile`s.
                     42: * Ability to generate HTML reports right form the base system, with such
                     43:   reports hooked into the various continuous build systems.
                     45: * Replacement of the `atf-run` and `atf-report` tools with shell scripts
                     46:   that provide backwards compatibility implementations based on `kyua`.
                     47:   These can deal with old-style `Atffile`s so that users with custom test
                     48:   suites can continue to run them.
                     50: * Introduction of a new `atf2kyua` script to convert old `Atffile`s to
                     51:   `Kyuafile`s.  This is used internally by the compatibility `atf-run`
                     52:   script, but can also be invoked by the end user by hand to deal with his
                     53:   own test suites.  Could be placed into `libexec` if we do not want to
                     54:   make this public.
                     56: ## For NetBSD 8.0
                     58: * Removal of the backwards compatibility `atf-run` and `atf-report`
                     59:   scripts, as well as the supporting `atf2kyua` tool.
                     61: # Why now?
                     63: Kyua has existed for almost 3 years already, so you may be wondering why
                     64: this import is being proposed now.  The major reasons are:
                     66: * Since 0.5, Kyua has now feature-parity with ATF.  Replacing the ATF tools
                     67:   with Kyua should not introduce functionality regressions.
                     69: * NetBSD-6 has been branched relatively recently, so we can expect NetBSD-7
                     70:   to be far away enough in the future to provide plenty of time to
                     71:   stabilize Kyua in NetBSD-current.
                     73: * The existence of the deprecated `atf-run` and `atf-report` tools hinders
                     74:   the development of new features that require changes to the inteface of
                     75:   ATF test programs.  As an example: a highly requested feature is the
                     76:   ability to change the timeout settings of the test programs (for the
                     77:   benefit of old, slow platforms); doing this in the ATF codebase is
                     78:   tricky.
                     80: * Some developers continue to add old-style tests to the tree.  While this
                     81:   is suboptimal and against the policy dictated by core@, Kyua brings in a
                     82:   mechanism to allow running these tests unmodified (i.e. without adding
                     83:   any ATF calls in their code).  This, in turn, will let such developers
                     84:   add their tests more easily, and thus increase the tests coverage.
                     86: # The plan
                     88: This section details what the transition plan looks like and an estimate
                     89: timeline.  As all things that depend on free time (opensource software
                     90: hacking, in this case), take the estimates with a grain of salt.
                     92: ## Discuss this plan in tech-userlevel@
                     94: If you are here it's possibly because a review request for this plan has
                     95: already been published and thus the plan has already begun.  If
                     96: not... well, hold off for a little bit ;-)
                     98: This plan will be sent to the tech-userlevel@ mailing list asking for
                     99: comments.  **Two weeks shall be allowed for initial discussion.** Depending
                    100: on the outcome, the plan and/or the software will need to be adjusted
                    101: (which in turn may require significant amounts of time not yet accounted
                    102: for).  I'll be optimistic for now.
                    104: ## Get core@ approval
                    106: By policy, the import of any new software component into src requires core@
                    107: approval.  Even though this will have been discussed at length in
                    108: tech-userlevel@ (as per the previous step), the final decider on this issue
                    109: will be core@.  The corollary of this is that, if no consensus can be
                    110: reached in tech-userlevel@ regarding this plan, core@ will be asked to come
                    111: up with a decision.
                    113: If core@ approves the plan, the next steps shall start immediately.  If
                    114: core@ disagrees, core@ will be asked to provide advice on the corrections
                    115: that should be made before the plan can be approved.
                    117: ## Import Kyua into src
1.2     ! jmmv      119: As the [[introductory page to Kyua|/kyua]] describes, Kyua has been
        !           120: available in pkgsrc for a while and can be readily installed and used to
        !           121: run the tests from `/usr/tests`.
1.1       jmmv      122: 
                    123: However, because ATF lives in src, and because NetBSD aims to provide the
                    124: best environment for testing "out of the box", Kyua should be imported into
                    125: src just like ATF was.  The major reasons for this, as have been explained
                    126: in the past, are: first, to allow any new deployment of NetBSD to be
                    127: validated right after installation and continuously afterwards; and,
                    128: second, to permit the execution of tests during development without having
                    129: to install any additional software.
                    131: The specific steps to perform this import are:
                    133: 1. Import Lutok into `external/bsd/lutok/`.  This is a shared library that
                    134:    wraps the native Lua C interface in C++.  Lutok was originally part of
                    135:    Kyua, and was split into its own package per the request of some users
                    136:    that found this component useful on its own.
                    138: 1. Import Kyua into `external/bsd/kyua-cli/`.  This yields a new kyua
                    139:    binary in `/usr/bin`, a lot of test programs in `/usr/tests/kyua-cli`
                    140:    (around 100) and some auxiliary files in `/usr/share/kyua`.
                    142: 1. Protect all products of Lutok and Kyua with the existing `MKATF` knob.
                    143:    If desired, these could instead be protected by a new and separate
                    144:    `MKKYUA` knob, but because the transition is only temporary, I do not
                    145:    think it's worth the effort.
                    147: 1. Update `` to generate `Kyuafile`s *in addition to*
                    148:    `Atffile`s.
                    150: There is no real need to do this import in a branch given that this import
                    151: only adds new functionality without touching existing stuff.
                    153: All the preparatory work for the import can be done offline (in about two
                    154: weeks at most, given that I have mot of this ready).  Aside from the code
                    155: changes, this will involve the validation of NetBSD/amd64, NetBSD/i386 and
                    156: NetBSD/macppc builds (which are the ports I have access to).  If you
                    157: consider that some other tricky architecture should be build-tested
                    158: (sparc64?), let me know and I'll include it in the list.
                    160: The submission step to CVS, once all the code changes are ready locally,
                    161: and any post-commit validation checks will take a few hours.  Any build
                    162: breakage should be addressed in a timely manner (and can possibly be
                    163: worked-around by setting `MKATF=no`).
                    165: ## Update documentation
1.2     ! jmmv      167: The [[Kyua: An introduction for NetBSD users|/kyua]] wiki page will be
1.1       jmmv      168: updated to explain how Kyua is bundled in NetBSD and how to use the bundled
                    169: version.
1.2     ! jmmv      171: The [[Creating atf-based tests for NetBSD src|/tutorials/atf]] wiki page
        !           172: will be updated to account for the differences in test programs execution
        !           173: with Kyua instead of ATF.
1.1       jmmv      174: 
                    175: The afterboot(8) and tests(7) manpages will be adjusted to mention `kyua`
                    176: instead of the ATF tools.
                    178: ## Adjust continuous testing systems to use Kyua
                    180: With `kyua` being part of the release sets, it is possible to adjust the
                    181: continuous test systems to make use of this tool in the test environments
                    182: without having to take any additional step.
                    184: I'll work with gson@ and pgoyette@ to adapt their continuous testing
                    185: machines to use the new built-in `kyua` binary instead of `atf-run` and
                    186: `atf-report`.  I'm planning to do the necessary work to change `anita`
                    187: myself, and I expect to help them deploy the changes to their own systems.
                    188: Because Kyua and ATF will cohexist in the base system at this point,
                    189: migrating the continuous testing systems to Kyua can happen at its own
                    190: peace.
                    192: It might happen that Kyua misses some little detail needed by these
                    193: systems.  In that case, this may require a new release of Kyua and a
                    194: reimport into the tree.  Incremental imports with new features are much
                    195: easier than the original import described in here.  Also, it will be
                    196: possible to cherry-pick any external changes into the tree without a
                    197: reimport (as has often been done in ATF).
                    199: ## User validation period
                    201: At this stage, **at least one month shall be given to the community** to
                    202: test the new tools and the new test results dashboards.  Collect feedback
                    203: and address requests as appropriate.
                    205: ## Replace atf-run and atf-report with kyua-atf-compat
                    207: Import the Kyua-based `atf-run` and `atf-report` compatibility tools and
                    208: stop building the deprecated versions of these.  The compatibility versions
                    209: are shipped in a separate `kyua-atf-compat` package, and thus this will be
                    210: imported into `external/bsd/kyua-atf-compat/`.
                    212: Once this is done, change `` to not generate `Atffile`s any
                    213: more, as the compatibility tools do not need them.
                    215: ## Drop atf-run and atf-report's code
                    217: Delete the native `atf-run` and `atf-report` utilities from the source tree
                    218: (and from the upstream repository).  This also gets rid of a lot of
                    219: supporting helper code, which makes the various ATF libraries leaner (and
                    220: therefore benefits all test programs!).
                    222: ## Get rid of some ATF wrappers as a proof of concept
                    224: There are some test programs (e.g. the ones in the ipf test suite) that are
                    225: not ATF-based.  To plug them into the NetBSD test suite, we have
                    226: implemented ATF-based wrappers that invoke the original plain tests.  This
                    227: is certainly suboptimal, as this level of indirection hinders readability
                    228: and makes development harder.  (In particular, it prevents some key
                    229: developers from contributing tests in the first place.)
                    231: Because Kyua has the ability to run these "plain" (non-ATF) test programs
                    232: directly, and because a bunch of developers are really looking forward to
                    233: this feature, I will convert a few tests to not include ATF wrappers as a
                    234: proof of concept of this feature.  The ipf tests are probably a good choice
                    235: for this.
                    237: ## Remove atf-run and atf-report compatibility tools
                    239: This is still really far away in the future (needs to happen after NetBSD
                    240: 7.0 is branched), but I'm listing it for completeness of the plan.  The
                    241: idea is to get rid of any ATF compatibility scripts by NetBSD 8.0.  This
                    242: gives end users a full release cycle (that of 7) to adapt to Kyua while at
                    243: the same time being able to use the old-style tools.
                    245: # Possible concerns
                    247: This section attempts to collect the list of possible concerns and/or
                    248: objections that may come up during the review, together with an attempt of
                    249: rebuttal from my side.
                    251: ## Increased compilation time
                    253: The addition of any new component to src increases the build time of the
                    254: whole system.  In the case of Kyua, this increase might be noticeable
                    255: *because of the large amount of test programs provided (roughly 100)*, all
                    256: of which are in C++.
                    258: Note that ATF also had a relatively large codebase and a bunch of tests
                    259: (although not as many).  The "trick" was that the majority of these tests
                    260: were written in shell, and as such they did not increase the build time by
                    261: much.  However, they significantly increased the run time of the test
                    262: suite, and they were less detailed (mostly integration tests, few unit
                    263: tests) than the Kyua tests.
                    265: In order to mitigate this issue, the build of all pieces of Kyua will be
                    266: protected by `MKATF` so that people allergic to C++ can avoid the whole
                    267: thing.  Even more, there are some other additional provisions described
                    268: below.
                    270: ## Introduction of more C++ code in base
                    272: First of all, why did I use C++?  To make the implementation simpler and
                    273: safer (the RAII programming pattern is really useful in avoiding memory and
                    274: resource leaks with minimal effort).  And C++ is part of the base system
                    275: and a supported language, so there was no reason not to do so.  But that's
                    276: not the point of this item: if you don't like C++, this is not going to
                    277: convince you otherwise.
                    279: It's true that, if we count the number of lines, Kyua brings in more C++
                    280: code than what will eventually be dropped by the removal of the ATF tools.
                    281: However, because ATF was also C++, the import of Kyua itself does not make
                    282: the situation significantly worse.
                    284: Additionally, the two compilers we can really use (GCC and LLVM) already
                    285: use, or will soon use, C++ in their code base.  It is unlikely that we will
                    286: be able to remove all C++ support from base anytime soon due to this, while
                    287: at the same time keeping support for all the ports that NetBSD has.
                    289: Long term, if the use of C++ proves to be a problem, there are a few things
                    290: that can be done to slowly get rid of it that have been floating my mind
                    291: recently.  The first is the rewrite of the performance-critical parts of
                    292: Kyua in plain C.  This would involve splitting the runtime engine of a
                    293: single test case in its own binary (which, for other reasons may be a good
                    294: idea on its own).  The second is the rewrite of most user-interface code in
                    295: Lua, which in itself would bring some extensibility advantages to the
                    296: program.  Both options will be investigated when the time permits.  In the
                    297: meantime, the replacement of ATF with Kyua does not make things worse; it
                    298: just changes one chunk of code with another.
                    300: ## No need for Lutok as a public library
                    302: Kyua depends on Lutok, which is a C++ interface to Lua.  The code of this
                    303: library was originally part of Kyua, but I split it into its own project
                    304: because some users asked for it.
                    306: If there is no desire to ship Lutok as a shared library in `/usr/lib`, we
                    307: can build Lutok as a static private library and link Kyua against it.
                    308: There is no need to install this as a shared library.

CVSweb for NetBSD wikisrc <> software: FreeBSD-CVSweb