[cmake-developers] [CMake] Random behavior of CTEST_USE_LAUNCHERS

[David Cole]

Well, the mixing of lines of output from parallel compiles was the
main reason log scraping didn't work well.

At least with the launchers, we'd be able to "scrape" from a single
output stream associated with a single build step and we wouldn't have
the mixing of lines problem anymore.

Nils has a point in that we are already analyzing the output
line-by-line in order to do custom warning suppression. It would be
awesome to make it better. (But would also probably require some
corresponding CDash changes...)

It's not going to be an "easy do it in a day" kind of thing, but this
is a good idea to keep on the back burner.

(Nils's main challenge at the moment, if I understand correctly, is to
find the ONE or TWO unsuppressed warnings in an output stream that
contains 100's of warnings... so that he can then suppress those as
well)


David


On Tue, Oct 16, 2012 at 3:27 PM, Brad King <brad.king at kitware.com> wrote:
> On 10/16/2012 03:09 PM, Nils Gladitz wrote:
>> Well what would be awesome is having the launcher extract individual
>> diagnostic messages.
>
> Limitations of that kind of "log scraping" was precisely the reason we
> created CTEST_USE_LAUNCHERS in the first place.  At most CTest should
> add markup CDash can use to highlight unsuppressed warnings.  That
> way failure does not drop anything, messages are just less prominent.
>
> -Brad
> --
>
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html
>
> Please keep messages on-topic and check the CMake FAQ at: http://www.cmake.org/Wiki/CMake_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/cgi-bin/mailman/listinfo/cmake-developers