Loading swank crashes Lisp

[Eric Timmons]

Loading the swank system during a SLIME session causes the Lisp process to die/become unresponsive (on at least SBCL 1.3.9 and CCL 1.11).

Steps to reproduce:

1. Start a SLIME session.
2. Load swank system.
3. Switch to *inferior-lisp* buffer and see that Lisp has crashed.

Problem seems to have been introduced in 3.1.7.27.

I install SLIME through MELPA. Crash is reproducible using both swank from Quicklisp and from the MELPA installed version.

SBCL errors are attached.

[Eric Timmons (daewok)]

CCL errors. Much smaller.

[Faré (fahree)]

Ouch. I reproduced and indeed the culprit is 22865e39b which has operation-forced constantly return t.

Apparently, swank-loader:init doesn't really want to be called with :delete t :reload t (anymore, if it ever did -- but my uncertain memories is that it used to work a few years back; I suspect recent package changes of having broken it).

The only useful behavior is therefore for operation-forced to constantly return nil, rather than constantly return t.

See
https://gitlab.common-lisp.net/asdf/asdf/merge_requests/33

Also, please get the SLIME authors to fix their broken code.

[Faré (fahree)]

We decided that this is a bug in SLIME (fixed in git master), that was abusing some obsolete ASDF internals, in a way that wasn't even following the once intended purpose of said internal function. "Not even wrong."

We "fixed" ASDF by removing the long deprecated internals that SLIME was abusing, operation-forced, in 3.1.7.28.