[armedbear-devel] Re: Bugs

[Mark Evenson]

On Oct 23, 2013, at 2:47, Faré <fahree at gmail.com> wrote:

> While testing on Windows, two tests revealed a bug in ABCL:
> test-nested-components.script test-system-pathnames.script
> Unsupported case in TRANSLATE-DIRECTORY-COMPONENTS.
> Is there a good way to get a useful backtrace? We used:
>    (let ((*debug-io* stream)) (top-level::backtrace-command count))
> and it yielded no output (count bound to nil, stream to the log file).

SYS:BACKTRACE is the official interface.  See examples of usages in SLIME’s `swank-abcl.lisp`, lines 351++.

If you [look at the implementation of TOP-LEVEL::BACKTRACE-COMMAND][1], you will see that it is merely interrogating the undocumented special EXT:*SAVED-BACKTRACE*, which is set when CL:INVOKE-DEBUGGER is triggered, so for examining the state of your program it is probably not that useful.

[1]: http://abcl.org/trac/browser/trunk/abcl/src/org/armedbear/lisp/top-level.lisp#L115

I can confirm your failure via the shell script entry point, but I had problems trying to run the tests interactively.  

I updated “~/work/asdf/“ to the latest rev from git, and invoked ‘make’ to concatenate ASDF into ‘asdf.lisp.

From an ABCL under SLIME started with `—-noinit` with working directory `asdf/test/`, evaluating the form 

    '(#.(load "script-support.lisp") #.(asdf-test::da) #.(load-asdf) #.(frob-packages) #.(load "test-nested-components”))

gives me the error that “test-nested-components” can’t be loaded.  Indeed there is nothing loadable there:

    CL-USER> (directory "/Users/evenson/work/asdf/test/test-nested-components*")
    (#P"/Users/evenson/work/asdf/test/test-nested-components.script"
     #P"/Users/evenson/work/asdf/test/test-nested-components-1.asd”)

What triggers the creation of of the loadable file?  Do I need to initialize something else to look at this interactively?

--

"A screaming comes across the sky.  It has happened before but there is nothing to compare to it now."