ABCL use of contrib/swank-repl.lisp fails

[easye]

@stassats recent work on SWANK-REPL contrib on Feb 15, 2023, viz. those commits through 5e8fc7c has borked ABCL.

@dkochmanski 's work on #760 has borked the use of the contrib/slime-repl.lisp with ABCL.

I'll try to figure out how to get this working, but it involves papering over the incompleteness of ABCL's Gray streams implementation with errors of the form:

The value #<SWANK/GRAY::SLIME-INPUT-STREAM {4486D34}> is not of type STREAM.

[easye]

@stassats had accepted the previous non-conservative consolidation of the ASDF and cl:load paths for loading SLIME, but asked that I inform/patch things when ABCL had gotten a better Gray/ANSI/Java streams abstraction. That work is still largely theoretical, unfortunately.

[stassats]

It would be a good idea to add ABCL to the CI.

[easye]

It would be a good idea to add ABCL to the CI.

I'll look into it as part of verifying this fix.

[easye]

Some background on ABCL's streams:

Design of lisp streams in ABCL

Description of current implementation rationale https://abcl.org/trac/wiki/DesignOfStreams.

[easye]

Whew: my patches to ABCL's Gray Streams implementation and our pretty printer (XP) have matured to the point that I intend to release abcl-1.9.2 to solve this problem. From this release on, ABCL should be a fully compliant provider of the implementations needed in the code in <file:swank/gray.lisp>.

• Add ABCL to the CI to ensure this stays fixed.

• Add conditionals to warn that the <file:contrib/swank-repl.lisp> contrib will no longer work on versions lesser than abcl-1.9.2 with the next release of SLIME

---

Comments, suggestions on ways to proceed from @stassats always welcome.

[easye]

One thing that remains broken in my usage is that of the "external" SLIME contrib for telling emacs to render ANSI control sequences <https://github.com/deadtrickster/slime-repl-ansi-color>. I would like to fix the remaining problems in ABCL in dealing with cl:stream IO which, I think, is a timing issue as ABCL tends to break up the ANSI control sequences in way that seems to confound Emacs. Stay tuned Bear fans…

[easye]

One thing that remains broken in my usage […]

I have determined that the problem here lies in the "external" slime-repl-ansi-color contrib, so this is no longer a blocker.

• Fix/merge in ANSI color sequences to SLIME

[easye]

As of armedbear/abcl@847fa63, the GRAY-STREAMS support in ABCL has been improved to the point that contrib/swank-repl.lisp now works.

I am in the process of releasing abcl-1.9.2 <armedbear/abcl#594> which will contain this fix.

[zacque0]

As of armedbear/abcl@847fa63, the GRAY-STREAMS support in ABCL has been improved to the point that contrib/swank-repl.lisp now works.

I am in the process of releasing abcl-1.9.2 <armedbear/abcl#594> which will contain this fix.

Thanks, can confirm that it works! Was facing the error just now. After compiling ABCL with the latest commit, it now works with SLIME.

My environment:
OS: Ubuntu 22.04.2 LTS x86_64
Java:

openjdk 17.0.7 2023-04-18
OpenJDK Runtime Environment Temurin-17.0.7+7 (build 17.0.7+7)
OpenJDK 64-Bit Server VM Temurin-17.0.7+7 (build 17.0.7+7, mixed mode, sharing)

Emacs: GNU Emacs 28.2
Slime: slime-20230605.1004 (from MELPA)

[easye]

Closed with abcl-1.9.2