On 1/27/10 Jan 27 -10:09 AM, Faré wrote:
Just telling that it failed isn't very useful, especially when others can't reproduce (painful with old SBCL, very expensive with ACL).
Can you attach a full log of the failures? Does ACL work better with old version of the test suite? I remember that a lot of those tests were failing on clisp at least.
Understood --- I will try to gather some information and put this on launchpad.
I will see about getting a less moldy copy of SBCL to test on. So these tests all pass for you? If this is a 1.0.28 peculiarity, I'm inclined to ignore it.
Query: how does one attach a full log of the failures? These shell scripts are not particularly forthcoming about belching up a backtrace or anything like that. They just print out a list of failing tests and exit. I could show you the full trace of what is printed when ACL 8.1 fails, and although it's a little longer than what I emailed, nothing additional looks in any way useful.
I will root around inside the tests. It seems to me that for debugging test failures, we should provide a mode where the QUIT-ON-ERROR function does not, in fact, quit on error.
I note, BTW, that there's no support in run-tests.sh for 64-bit CCL, lispworks, or ABCL. OK, new launchpad ticket coming...
PS: I see you were in Cambridge MA recently. Next time you are, contact me!
Will do. I tried tweeting (I don't know a protocol for "tweeting" onto IRC), but that seems... suboptimal....
Next time I want to synchronize with the user group, too!
[ François-René ÐVB Rideau | Reflection&Cybernethics | http://fare.tunes.org ] Death is only a milestone - albeit one that is dropped on you from a very great height — Terry Pratchett.
2010/1/27 Robert Goldman rpgoldman@sift.info:
On 1/27/10 Jan 27 -12:50 AM, Faré wrote:
I've just released ASDF 1.501 in the official repository, now with all the source registry configuration that I previously discussed. It's currently documented in its own file README.source-registry, rather than in the general manual asdf.texinfo, as it should be. Patch welcome.
Note that I bumped the version from 1.375 to 1.500 then 1.501. This to indicate that we're not using CVS anymore, that I've reached a milestone towards my goal of an "ASDF 2" that I could push as a replacement to ASDF. It passes the tests with SBCL. But the tests could be extended to do more.
Next, comes a similar revamp of ASDF-BINARY-LOCATIONS configuration. Or maybe a wholesale replacement of ABL with something that's simpler and configured in a way similar to source-registry? What do YOU think?
I have an old copy of SBCL, 1.0.28, which I keep around (we pinned ourselves to that revision for a project I was working on), and I tried to run the test suite on this version of SBCL, 64-bit Mac.
The test suite failed, and here are the last several lines of the output:
; compilation unit finished ; caught 2 STYLE-WARNING conditions ; printed 1 note
; /Users/rpg/lisp/asdf/asdf.fasl written ; compilation finished in 0:00:07.450 Testuite failed: ASDF compiled with warningsbash-3.2$
I thought that this might be a spurious failure having to do with being too stringent about what constitutes an ASDF compilation failure, so I tried to run the test suite again (figuring a compiled version of asdf.lisp would now be available), but it failed identically.
Is this expected? Should I ticket this?
I will report on ACL tests shortly.
Thanks, Robert