Ran 35 tests: 35 passing and 0 failing all tests apparently successful
Using alisp -q -batch Ran 35 tests: 35 passing and 0 failing all tests apparently successful
Using mlisp -q -batch Ran 35 tests: 35 passing and 0 failing all tests apparently successful
Using /Users/rpg/ccl/dx86cl64 --no-init --quiet --batch Ran 35 tests: 35 passing and 0 failing all tests apparently successful
Using clisp -norc -ansi -I -on-error exit Ran 35 tests: 35 passing and 0 failing all tests apparently successful
Using ecl -norc -load sys:cmp Ran 35 tests: 34 passing and 1 failing failing test(s): run-shell-command-test.script
Using sbcl --noinform --userinit /dev/null --sysinit /dev/null --disable-debugger Ran 35 tests: 35 passing and 0 failing all tests apparently successful
I.e., only ECL is failing any of the scripts. That's actually more a bug with my test --- ECL will never be able to capture STDERR. Attached is a proposed test modification.
r
P.S. I have a dead-simple perl script that cycles through a configurable set of lisp impementations, running all the tests. If I can get it tidied, I will make it publicly available.
On Fri, Oct 21, 2011 at 15:56, Robert Goldman rpgoldman@sift.info wrote:
I.e., only ECL is failing any of the scripts. That's actually more a bug with my test --- ECL will never be able to capture STDERR. Attached is a proposed test modification.
That modification was already included in 07ad929 (right after 2.017.21).
P.S. I have a dead-simple perl script that cycles through a configurable set of lisp impementations, running all the tests. If I can get it tidied, I will make it publicly available.
Is it better than make test-all lisps="ccl clisp sbcl ecl cmucl abcl scl allegro" ?
—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
On 10/21/11 Oct 21 -3:56 PM, Faré wrote:
On Fri, Oct 21, 2011 at 15:56, Robert Goldman rpgoldman@sift.info wrote:
I.e., only ECL is failing any of the scripts. That's actually more a bug with my test --- ECL will never be able to capture STDERR. Attached is a proposed test modification.
That modification was already included in 07ad929 (right after 2.017.21).
P.S. I have a dead-simple perl script that cycles through a configurable set of lisp impementations, running all the tests. If I can get it tidied, I will make it publicly available.
Is it better than make test-all lisps="ccl clisp sbcl ecl cmucl abcl scl allegro"
Not sure. I am not fully trained on your make scripts.
The only possible betterness is that it parses out the blocks that say what the results are, so that you don't have to read through the full clutter (which is available in a /tmp file). You saw what the script drops to standard output.
Question: would it be reasonable to change the makefile to something like
lisps ?= ${ASDF-TEST-LISPS}
(I may well have the syntax wrong here)
so that people can do
export ASDF-TEST-LISPS "ccl clisp sbcl ecl cmucl abcl allegro allegromodern"
in their shell configuration?
cheers, r
The only possible betterness is that it parses out the blocks that say what the results are, so that you don't have to read through the full clutter (which is available in a /tmp file). You saw what the script drops to standard output.
The current `make test-all` stops at the first failing Lisp implementation. Not great, but at least makes failure obvious.
Question: would it be reasonable to change the makefile to something like
lisps ?= ${ASDF-TEST-LISPS}
(I may well have the syntax wrong here)
so that people can do
export ASDF-TEST-LISPS "ccl clisp sbcl ecl cmucl abcl allegro allegromodern"
in their shell configuration?
Wish granted.
—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
On 10/22/11 Oct 22 -2:09 PM, Faré wrote:
The only possible betterness is that it parses out the blocks that say what the results are, so that you don't have to read through the full clutter (which is available in a /tmp file). You saw what the script drops to standard output.
The current `make test-all` stops at the first failing Lisp implementation. Not great, but at least makes failure obvious.
Question: would it be reasonable to change the makefile to something like
lisps ?= ${ASDF-TEST-LISPS}
(I may well have the syntax wrong here)
so that people can do
export ASDF-TEST-LISPS "ccl clisp sbcl ecl cmucl abcl allegro allegromodern"
in their shell configuration?
Wish granted.
Hm. This seems to behave differently from simple run-tests. E.g., run-tests works fine for me, finding CCL from environment variables, but 'make-test-all-lisps` fails, being unable to find ccl...
The problem is in test-upgrade, invoked by test-all-lisps.
run-tests.sh ccl does:
command="${CCL:-ccl}"
but test-upgrades does:
use_ccl () { li="ccl --quiet --load test/script-support" ; ev="--eval" ; } ; \
Also, I find that for some lisps test-upgrades mistakenly loads my lisp init, causing spurious fails in some cases.
Makefile patch attached.
Hm. This seems to behave differently from simple run-tests. E.g., run-tests works fine for me, finding CCL from environment variables, but 'make-test-all-lisps` fails, being unable to find ccl...
The problem is in test-upgrade, invoked by test-all-lisps.
run-tests.sh ccl does:
command="${CCL:-ccl}"
but test-upgrades does:
use_ccl () { li="ccl --quiet --load test/script-support" ; ev="--eval" ; } ; \
Also, I find that for some lisps test-upgrades mistakenly loads my lisp init, causing spurious fails in some cases.
Makefile patch attached.
I updated the Makefile based on your suggestion, but with various simplifications. See 4bdcc680
—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
I figured you would do it over better ;-)
-
Faré -
Robert Goldman -
Robert P. Goldman