When I configure with a prefix, build and install, 95 '.dump' files are
installed (see
https://github.com/parrot/parrot/issues/512#issuecomment-5958144). Why?
Those '.dump' files are Perl Storable files -- a file format not
readable by Parrot. If they're not readable, they would seem to be of
no subsquent use to the installed Parrot.
Here's a use case. Suppose that I make a list of all the test files
under t/ which have a '#!./parrot' shebang line. I can run all those
files with the build *parrot* with satisfactory results:
find t -type f -name '*.t' | xargs grep -lE '#!\.\/parrot$' | xargs
-i ./parrot {}
I can also run those files with the *installed* parrot:
find t -type f -name '*.t' | xargs grep -lE '#!\.\/parrot$' | xargs
-i /home/jimk/work/pseudoinstall/bin/parrot {}
The tests continue to pass.
Now here's the interesting part. With the exception of vtable.dump, all
those '.dump' files are installed under ./src/parrot/4.4.0-devel/.
(vtable.dump sits at the top level of Parrot -- ./ in this case.)
Suppose I move all those file out of the way:
mv -v ./src/parrot/4.4.0-devel/*.dump ~/tmp/dumps/
mv -v ./vtable.dump ~/tmp/
Now suppose I re-run all the tests written in Parrot with the installed
Parrot, just as I did above.
All the tests continue to pass. That implies either (a) the installed
./dump files are useless and don't need to be installed; or (b) the
portion of our current test suite written in Parrot has a major deficit
in that it fails to test those aspects of Parrot which do, somehow, use
those .dump files.
Is my reasoning correct? Is there something I'm missing? If the
installed .dump files are useless, then is there any need to convert
them to a format such as JSON, as suggested in Issue #512?
Thank you very much.
Jim Keenan
_______________________________________________
http://lists.parrot.org/mailman/listinfo/parrot-dev
