OK, I won't be doing this. I didn't pay enough attention: this requires that I build clisp. With some exceptions (SBCL, mkcl), I have decided not to build lisp implementations as part of the ASDF testing process. The process is already too burdensome as it is. If clisp's own maintainers cannot make a release version, I can't be filling that hole myself -- I'm just a humble ASDF maintainer, not The Debugger of All Lisp Implementations.

At some point, clisp will either make a new release, or I will formally un-support it, and remove it from my testing routines. In the shorter term, I will probably stop running test-program.script on clisp.

Of course, I will always accept patches that clisp needs from others, but I don't have the personal resources to fill this hole.

On 28 Jul 2018, at 15:39, Faré wrote:

CLISP has long had semi-deterministic SIGSEGV issues like that:
reproducible, but going away at the slightest change in the test or
its invocation made trying to debug it. I suspect a GC bug. The CLISP
maintainers asked me to run the tests with some debugging option
enabled (I don't remember which), but of course, the change makes the
problem disappear. If you use Jenkins, maybe you should include a
build with those options so the bug can be tracked when it happens
(which is not at the same time as when it happens without the
debugging options).

See https://sourceforge.net/p/clisp/bugs/678/

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
There is no meaning without words, no shape without matter,
no soul without body.

On Sat, Jul 28, 2018 at 4:28 PM Robert Goldman rpgoldman@sift.info wrote:

For some reason, the ASDF tests are failing on Jenkins on linux with clisp, but only when run in Jenkins and only for clisp. At the command line, all the tests pass for me.

Under Jenkins, I get this failure on test-program.script:

;; Loading file test-program.script ...
FOO
(UIOP/LISP-BUILD:CURRENT-LISP-FILE-PATHNAME) => #P"test-program.script"
; Registering system hello-world-example
0 errors, 0 warnings
;; Loading file /var/lib/jenkins/workspace/asdf/test/hello-world-example.asd ...
;; Loaded file /var/lib/jenkins/workspace/asdf/test/hello-world-example.asd
0 errors, 0 warnings
0 errors, 0 warnings
; Registering system lisp-invocation
;; Loading file /var/lib/jenkins/workspace/asdf/asdf.asd ...
;; Loaded file /var/lib/jenkins/workspace/asdf/asdf.asd
0 errors, 0 warnings
;; Loading file /var/lib/jenkins/workspace/asdf/build/fasls/clisp-2.49.60+-unix-x64/asdf/build/asdf.fas ...
;; Loaded file /var/lib/jenkins/workspace/asdf/build/fasls/clisp-2.49.60+-unix-x64/asdf/build/asdf.fas
0 errors, 326 warnings
;; Loading file /var/lib/jenkins/workspace/asdf/ext/lisp-invocation/lisp-invocation.asd ...
;; Loaded file /var/lib/jenkins/workspace/asdf/ext/lisp-invocation/lisp-invocation.asd
0 errors, 0 warnings
0 errors, 0 warnings
0 errors, 0 warnings
;; Loading file /var/lib/jenkins/workspace/asdf/ext/lisp-invocation/lisp-invocation.asd ...
;; Loaded file /var/lib/jenkins/workspace/asdf/ext/lisp-invocation/lisp-invocation.asd
0 errors, 0 warnings
0 errors, 0 warnings
0 errors, 0 warnings
0 errors, 0 warnings
;; Loading file /var/lib/jenkins/workspace/asdf/ext/fare-utils/fare-utils.asd ...
;; Loaded file /var/lib/jenkins/workspace/asdf/ext/fare-utils/fare-utils.asd
0 errors, 0 warnings
*** - handle_fault error2 ! address = 0x55a16f64c670 not in [0x5575be29b000,0x5575be657030) !
SIGSEGV cannot be cured. Fault address = 0x55a16f64c670.
GC count: 65
Space collected by GC: 74922832
Run time: 0 874720
Real time: 1 803489
GC time: 0 179953
Permanently allocated: 165368 bytes.
Currently in use: 5850000 bytes.
Free space: 1462440 bytes.
Segmentation fault (core dumped)
Using clisp, test-program.script failed

This is with the latest release (but ancient) clisp.

It looks like it's not even making it through the preliminary loading process, since we don't see the "test image-op" debugging output.

My guess is that this is not a real asdf failure, but instead is clisp being unable to find some piece of tool-age that it needs in order to build a program.

But this happened only after (1) load-asdf changed and (2) source-file-type was deprecated. Neither of those seems obviously connected to this issue.

Any hypotheses or suggestions for debugging?