[slime-devel] Re: sbcl regression with slime

[Christophe Rhodes]

Alexey Dejneka <adejneka at comail.ru> writes:

> Zach Beane <xach at xach.com> writes:
>
>> To reproduce:
>>
>>    1. Start SLIME with a recent CVS sbcl
>>
>>    2. At the slime repl, type "*features*"
>>
>> At that point, I get a message like "error in process filter: wrong
>> type argument listp," in the minibuffer. (When I try to select it, it
>> disappears, sorry for the imprecision.)
>
> I had the same problem when I updated SBCL to the Unicode version. But
> for me removing .slime/sbcl*/*.fasl and restarting emacs has solved
> the problem. Just updated SLIME and SBCL 0.8.16.35 with :SB-UNICODE
> pass your test here.

How odd.

To me, it seems that this bug has always been there, and I have no
idea why we've only just started noticing it.  (Clearly I don't do
enough real work in slime).  The problem arises in UTF-8 locales only,
with slime's emacs end treating the first three bytes or so of a
connection as text (rather than binary).

The attached patch fixes the problem at a certain cost in dynamic
range of communications.  Please advise if this is the desired
solution, or if I/we need to do some more work.

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: slime.diff
URL: <https://mailman.common-lisp.net/pipermail/slime-devel/attachments/20041109/51ac3901/attachment.ksh>
-------------- next part --------------

Cheers,

Christophe