r207951 - Update lit.cfg to support the clang-interpreter test from r207950

[Richard Smith]

On Sat, May 31, 2014 at 7:18 PM, Alp Toker <alp at nuanti.com> wrote:

>
> On 01/06/2014 04:54, Richard Smith wrote:
>
>> On Sat, May 31, 2014 at 6:46 PM, Alp Toker <alp at nuanti.com <mailto:
>> alp at nuanti.com>> wrote:
>>
>>
>>     On 01/06/2014 04:05, Richard Smith wrote:
>>
>>         On Sun, May 4, 2014 at 11:42 PM, Alp Toker <alp at nuanti.com
>>         <mailto:alp at nuanti.com> <mailto:alp at nuanti.com
>>
>>         <mailto:alp at nuanti.com>>> wrote:
>>
>>             Author: alp
>>             Date: Mon May  5 01:42:07 2014
>>             New Revision: 207951
>>
>>             URL: http://llvm.org/viewvc/llvm-project?rev=207951&view=rev
>>             Log:
>>             Update lit.cfg to support the clang-interpreter test from
>>         r207950
>>
>>             Performs behind-the-scenes RUN line substitution similarly to
>>             what's done with
>>             clang-check and clang-format to ensure the executable is
>>         found.
>>
>>             Modified:
>>                 cfe/trunk/test/CMakeLists.txt
>>                 cfe/trunk/test/lit.cfg
>>
>>             Modified: cfe/trunk/test/CMakeLists.txt
>>             URL:
>>         http://llvm.org/viewvc/llvm-project/cfe/trunk/test/
>> CMakeLists.txt?rev=207951&r1=207950&r2=207951&view=diff
>>                    ==============================
>> ================================================
>>             --- cfe/trunk/test/CMakeLists.txt (original)
>>             +++ cfe/trunk/test/CMakeLists.txt Mon May  5 01:42:07 2014
>>             @@ -29,6 +29,7 @@ list(APPEND CLANG_TEST_DEPS
>>                c-index-test diagtool arcmt-test c-arcmt-test
>>                clang-check clang-format
>>                clang-tblgen
>>             +  clang-interpreter
>>
>>
>>         I'm all in favor of having buildbots check that the
>>         clang-interpreter example still works, but is it really
>>         sufficiently valuable to slow down *all* clang-test builds by
>>         adding another (slow) link action? Can we make this only run
>>         on the buildbots by default somehow?
>>
>>
>>     They aren't run by default. Perhaps you're passing
>>     -DCLANG_BUILD_EXAMPLES=1 which causes the examples to get built
>>     and included in the test set?
>>
>>
>> Nope.
>>
>> $ grep CLANG_BUILD_EXAMPLES build/CMakeCache.txt
>> CLANG_BUILD_EXAMPLES:BOOL=OFF
>>
>> Note that your change above adds clang-interpreter unconditionally to
>> CLANG_TEST_DEPS.
>>
>
> Fixed in r209983! However..
>
>
>
>>     I don't think it makes sense to have an additional mode that
>>     enables the examples but excludes their tests.
>>
>>     I've just timed the test deps build with and without all examples
>>     on an old laptop and the difference isn't measurable above noise:
>>
>>     With examples and example tests enabled:
>>
>>     touch ../upstream/clang/lib/Basic/Diagnostic.cpp && time ninja
>>     check-clang-deps
>>     real    0m1.181s
>>     user    0m2.648s
>>     sys    0m1.061s
>>
>>     Without examples or example tests enabled:
>>
>>     touch ../upstream/clang/lib/Basic/Diagnostic.cpp && time ninja
>>     check-clang-deps
>>     real    0m1.283s
>>     user    0m2.637s
>>     sys    0m1.065s
>>
>>     Beyond that, the complete run time for all the example tests
>>     combined looks like it weighs in at ~1/20th of a second in a full
>>     test run.
>>
>>
>> The problem is the link time, not the runtime of the tests.
>>
>
> I don't see how this could have affected link time during development as
> the timings above demonstrate.
>

Your before and after timings both include linking the binaries, don't they?


> (Unless you're using LTO or static builds to develop,


I don't know anyone who uses anything *other* than static builds. That's
our default and recommended build configuration.


> which will have all kinds of overheads -- but why would anyone do that?)
>
> Alp.
>
>
>
>
>>     Surely all of this is insignificant when there are tests that
>>     generate large files or sleep for seconds? Also consider using lit
>>     directly with the --incremental flag.
>>
>>     Alp.
>>
>>
>>
>>                PrintFunctionNames
>>                SampleAnalyzerPlugin
>>                )
>>
>>             Modified: cfe/trunk/test/lit.cfg
>>             URL:
>>         http://llvm.org/viewvc/llvm-project/cfe/trunk/test/lit.
>> cfg?rev=207951&r1=207950&r2=207951&view=diff
>>                    ==============================
>> ================================================
>>             --- cfe/trunk/test/lit.cfg (original)
>>             +++ cfe/trunk/test/lit.cfg Mon May  5 01:42:07 2014
>>             @@ -308,6 +308,7 @@ for pattern in [r"\bFileCheck\b",
>>                              r"\bc-index-test\b",
>>                              NoPreHyphenDot + r"\bclang-check\b" +
>>             NoPostHyphenDot,
>>                              NoPreHyphenDot + r"\bclang-format\b" +
>>             NoPostHyphenDot,
>>             +                NoPreHyphenDot + r"\bclang-interpreter\b" +
>>             NoPostHyphenDot,
>>                              # FIXME: Some clang test uses opt?
>>                              NoPreHyphenDot + r"\bopt\b" +
>>         NoPostHyphenDot,
>>                              # Handle these specially as they are
>>         strings searched
>>
>>
>>             _______________________________________________
>>             cfe-commits mailing list
>>         cfe-commits at cs.uiuc.edu <mailto:cfe-commits at cs.uiuc.edu>
>>         <mailto:cfe-commits at cs.uiuc.edu <mailto:cfe-commits at cs.uiuc.edu>>
>>
>>         http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>
>>
>>
>>     --     http://www.nuanti.com
>>     the browser experts
>>
>>
>>
> --
> http://www.nuanti.com
> the browser experts
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140531/0180907a/attachment.html>