Michał "phoe" Herda
Michał "phoe" Herda
[The job](https://travis-ci.com/phoe-trash/roswell-test/jobs/272523026) failed with a segfault. ``` /c/Users/travis/.travis/functions: line 109: 1413 Segmentation fault ros init C:/Users/travis/build/phoe-trash/roswell-test/roswell/ros.exe: error while loading shared libraries: ?: cannot open shared object file: No such file...
Fails on standalone Windows as well with `zlib1.dll` not found on the system. 
```lisp (defun test () (let ((var 0)) (declare (type (and number (real (0))) var)) (+ 23 var))) ``` This is undefined behaviour. You tell the compiler that the value `0`...
Thank you. If I understand this context correctly now, CCL `LOOP` is buggy, as in, it expands into code which invokes undefined behaviour.
OK, thanks. Sorry, I shouldn't perhaps post when exhausted. This is not a spec bug in CCL per se, because the code wasn't conforming in the beginning. If anything, a...
> If a macroexpansion-time error is kosher That's a question to @xrme - we are already in undefined behaviour territory, so CCL is free to do whatever it wants to....
@kpoeck Thanks for the fix! Would you also mind submitting a matching regression test case to https://github.com/Clozure/ccl-tests/ that shows that the bug has been fixed?
I encountered that issue in #204 and had to work around by making several more building steps - can you check if that workaround works for you? On 13.11.2021 01:33,...
I think that there is some sort of intermediate commit that you must checkout in order to work around the bug that prevents `optimizers.lisp` from getting compiled. I think it...
Brief analysis: The function `%MAYBE-STD-SETF-SLOT-VALUE-USING-CLASS` mentioned in the stack trace is: https://github.com/Clozure/ccl/blob/275105afd94706d95ac955178316074931822c42/level-1/l1-clos-boot.lisp#L2836-L2851 It calls an inlined function named `%SET-STD-SLOT-VECTOR-VALUE`: https://github.com/Clozure/ccl/blob/275105afd94706d95ac955178316074931822c42/level-1/l1-clos-boot.lisp#L2805-L2823 The `BAD-SLOT-TYPE` error is signaled from here. Likely `TYPE` and...