I noticed today that Werner had not updated the list of D examples for tests
so "make test_noninteractive" (and ctest) were giving an impression of D that
was worse than it actually is.  I have now fixed that oversight and the
results are as follows:

d
   Missing examples            :  14 14a 20 21 22 28 31
   Differing postscript output :  02 12 13 17
   Missing stdout              :
   Differing stdout            :  01

I had to exclude the 14th example because of a run-time error caused by
plgdev not working correctly for the D bindings.  Example 1 gives the wrong
stdout because plgver is not working correctly.  I haven't investigated the
issues for the 4 examples with differing PostScript results.  The above
results are better than previous ones and are quite encouraging.

As of revision 10082, I implemented convenient options for the new build
system for the installed examples that allow the user to cut down on the
examples that are tested and run to save test time and disk space.

Here are the commands I used to obtain the above result:

softw...@raven> pwd
/home/software/plplot_install_test/build_dir
softw...@raven> rm -rf /home/software/plplot_install_test/build_dir/*
softw...@raven> cmake -DDEFAULT_NO_BINDINGS=ON -DENABLE_d=ON \
-DDEFAULT_NO_DEVICES=ON -DPLD_psc=ON \
~/plplot_cvs/installcmake/share/plplot5.9.4/examples >& cmake.out
softw...@raven> less cmake.out
softw...@raven> time make -j4 test_noninteractive >& make_test.out

real    0m18.562s
user    0m10.397s
sys     0m2.900s
softw...@raven> less make_test.out
softw...@raven> time make -j4 test_noninteractive >& make_test.out1

real    0m0.631s
user    0m0.564s
sys     0m0.428s

To put these timing results in context, I only have an entry level desktop
PC (Intel duo core, 2.4GHz).

The first timing test is a bit of a cheat because D example 20 has not
been implemented yet, and the 20th example always takes a substantial fraction
of the total time for the complete run of examples.  Nevertheless, only ~20
seconds from scratch is excellent for building the C and D examples,
running all of them that have been implemented, and comparing the results.

The second timing test shows there is low latency in the results if there is
nothing that has been updated.  The latency goes up to something like 6
seconds for a fully configured complete test_noninteractive, but that is
expected.  That larger latency reflects the huge number of dependencies that
are being checked in that case to figure out nothing has to be updated.
(Indeed, I double-checked that nothing was updated.)

Anyhow, these latency tests are a good indication that I haven't messed up
any dependencies in the new build system.  (That was not the case before I
starting working on the new build system today to make it more convenient to
test the D results.)

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

------------------------------------------------------------------------------
_______________________________________________
Plplot-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/plplot-devel

Reply via email to