On 6/23/15 Jun 23 -1:46 PM, Dave Cooper wrote:
You have to set that ALLEGRO_NOISY variable to true (in my run script). That will cause allegro to show its console, which is where you will see most of the output on Windows.
It is controlled by the +c argument.
Right. So this variable is off by default when running the test script.
Thanks. When I turn this on, I can see where we went awry.
But... on Unix-like systems where there's no console, the output of the tests goes into the test script, so that you can see exactly what comparisons failed. That's obviously much more desirable than this behavior.
Is there no way to redirect the output of ACL on Windows? Must it always go into its own console window (or be lost altogether if it's run without a console window)? That seems deeply undesirable....
Best, r
On Tuesday, June 23, 2015, Robert Goldman <rpgoldman@sift.net mailto:rpgoldman@sift.net> wrote:
I'm having some real oddities testing on Windows. Attached is a roster of the implementations for which I see test failures -- the Allegro variants and clisp. Two oddities: 1. I do not get verbose output showing what happens when the tests fail. Critical information is missing from the logfiles. I suspect that the test script assumes that standard output is redirectable in some way that it is not on Windows. See allegro-test.txt which is attached. Any idea how to get the lisp output? Any chance that REDIRECT-OUTPUTS function from script-support.lisp does not work on Allegro + Windows? 2. I get fewer failures running one test at a time than running them all in make. Allegro testing reported the following: -#--------------------------------------- Using g:/windows/allegro/32/non-smp/alisp Ran 56 tests: 52 passing and 4 failing failing test(s): test-encodings.script test-program.script test-try-refinding.script test-utilities.script -#--------------------------------------- But when I run the tests individually, only test-program and test-utilities failed. The other two tests succeeded. I have rarely seen something like this before in make on Unix variants due to not cleaning up the filesystem, but never in such a wholesale way. ---- Trying not to be a Unix bigot, but failing.... Best, R-- My Best,
Dave Cooper genworks.com http://genworks.com, gendl.org http://gendl.org +1 248-330-2979