[Ecls-list] [cl-test-grid] cl-test-grid incremental reports

[Anton Vodonosov]

Meantime I've tested 89a8201b and compared with the "base version" b6519e5c.

Now we have some improvements, but also regressions and "infrastructure problems"

bytecode compiler: http://common-lisp.net/project/cl-test-grid/ecl/ecl-diff-bytecode.html

  - genworks-gdl timeouts are not regressions.
    It's just that bytecode compiler really takes > 10 minutes to install genworks-gdl.
    Tried with the 12.12.1 release, it works the same.
    I have impression that the time is spent during unpacking of the 20 MB .tgz archive.
    The timeout didn't happened in previous testing because genworkd-gdl has already been
    downloaded by some other lisp implementation. But now bytecode ecl
    was the first lisp implementation run on this quicklisp so the download was performed
    by it.
  - opticl-doc not a regression.
    Again, we see it this time, but not in previous test runs because this time
    bytecode ecl was the first lisp running (ql:quickload :opticl-doc).
 
    opticl-doc generates .html file from .md file during asdf:load-op,
    and signals an error during this process. Nevertheless, the .html file
    is created, and subsequent asdf:load-ops just do nothing, therefore
    the error only signaled by first asdf:load-op.

lisp-to-c compiler: http://common-lisp.net/project/cl-test-grid/ecl/ecl-diff-lisp-to-c.html
  - com.informatimago.common-lisp systems - real regression, confirmed manually
  - gbbopen - not a regression. This system, gbbopen, doesn't use ASDF but
    uses it's own module manage, and stores .fas files in a subdirectory along the source
    code. Therefore the asdf-output-translations installed by test-grid-agent do not
    work and the .fas files are not recompiled during every test run, but old .fas files
    are used left there from previous runs. In this case the .fas file was left here
    by ecl bytecode compiler. The .fas file is strange indeed, and you can
    reproduce the " invalid ELF header" problem during bytecode compilation by:
       rm -r quicklisp/dists/quicklisp/software/gbbopen-20121223-svn/linux86-ecl-12.12.1/
       lisps/ecl-bin/bin/ecl -eval "(ext:install-bytecodes-compiler)"  -eval "(ql:quickload :gbbopen)" -eval "(quit)"
    But this is definitely not a regression, it happens on older versions too.
  - lucene-in-action-tests and montezuma - real regression, confirmed manually


So, in total 2 new regressions on lisp-to-c compiler, and some problems which can
not be qualified as regressions, which you may or may not want to investigate.

As we have regressions, "base version" is not shifted forward, it remains at b6519e5c.

Best regards,
- Anton