[Ecls-list] strange bug related to streams

Stelian Ionescu sionescu at common-lisp.net
Sat Feb 2 22:49:51 UTC 2008

On Sat, Feb 02, 2008 at 10:47:37PM +0100, Juan Jose Garcia-Ripoll wrote:
>On Feb 2, 2008 10:18 PM, Stelian Ionescu <sionescu at common-lisp.net> wrote:
>> I've encountered a strange bug while trying to load CFFI:
>> ;;; Compiling (DEFUN %DEFCFUN-VARARGS ...).
>> Cannot unread the stream #<string-input stream from "-">. [...]
>> the problem is caused by #\- (and other cases of literal characters)
>Hi Stelian,
>could you be more precise about the steps to reproduce this bug? I
>have just downloaded CFFI, and used (asdf:oos 'asdf:load-op :cffi)
>from inside the source directory, and everything works just fine.

The latest release(0.9.2) compiles well here, too

>The error message you just showed could be caused by an improper use
>of unread-char, such as unreading a charcter that has not been read
>before. I need more information to track this issue.

I'm using the development version of CFFI. Here's a list of all repos
you have to download to make it work(all Darcs):


Also I'm using ASDF-BINARY-LOCATIONS to separate fasls from sources(it
generally works better than COMMON-LISP-CONTROLLER)
Finally, there's my ~/.eclrc which does little more than loading
ASDF-BINARY-LOCATIONS and creating my own -user package with a few
functions in it, nothing fancy

I've just noticed another interesting behaviour: compiling SWANK(SLIME's
CL side) works well, but if I try to compile it *after* I got that error
with CFFI, compiling swank-backend.lisp signals
Cannot unread the stream #<string-input stream from "s">.
the s being that of a #\space in GUESS-EXTERNAL-FORMAT

Stelian Ionescu a.k.a. fe[nl]ix
Quidquid latine dictum sit, altum videtur.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <https://mailman.common-lisp.net/pipermail/ecl-devel/attachments/20080202/53aef61e/attachment.sig>

More information about the ecl-devel mailing list