Guile v. 2.0.1 does not handle escaped open parentheses correctly.

Guile v. 2.0.1 does not handle escaped open parentheses correctly.

[Torstein Winterseth]

Hello,

I have found a bug where Guile does not handle escaped open
parentheses correctly. This was found in a real program which works in
Dr. Racket in R5RS mode. You are not the only interpreter on Fedora 15
which fails this, but anyhow.

Testcase:

(define bug
 '(\())

Expected results:
No errors. bug variable should now be bound to the list (().

Actual result:
[torstein@Desktor Skrivebord]$ guile -l testcase.scm  --no-auto-compile
ERROR: In procedure primitive-load:
ERROR: In procedure scm_i_lreadparen:
/home/torstein/Skrivebord/testcase.scm:3:12: end of file

In the actual program the list was larger, '(\{ \[ \(), and then the error was:
ERROR: In procedure primitive-load:
ERROR: In procedure scm_read_expression:
/home/torstein/Skrivebord/testcase.scm:2:27: unexpected ")"

-- 
Mvh / Kind regards
Torstein

Re: Guile v. 2.0.1 does not handle escaped open parentheses correctly.

[Andy Wingo]

Hi Torstein,

On Wed 11 May 2011 16:22, Torstein Winterseth <kvikende@fsfe.org> writes:

> (define bug
>  '(\())

AFAIU, \ is not actually an escaping marker in R5RS.  The meaning of
your program is not defined by the R5RS, as an initial \ character on a
form is not specified.  Guile treats this the same as:

However...

  $ guile
  scheme@(guile-user)> '\(
  $1 = \
  scheme@(guile-user)> 1
  ... )
  <unnamed port>:0:0: In procedure #<procedure 18d0500 at <current input>:1:0 ()>:
  <unnamed port>:0:0: Wrong type to apply: 1

  Entering a new prompt.  Type `,bt' for a backtrace or `,q' to continue.
  scheme@(guile-user) [1]> 

Not sure why it decided that it should display a prompt after the first
case; probably a bug somewhere...

But, your assumption is not correct, so this is not a bug, AFAIK.  R6RS
does specify an escaping syntax, however, though it is not the one you
are assuming.

Regards,

Andy
-- 
http://wingolog.org/

Re: Guile v. 2.0.1 does not handle escaped open parentheses correctly.

[Ian Price]

Torstein Winterseth <kvikende@fsfe.org> writes:

> Hello,
>
> I have found a bug where Guile does not handle escaped open
> parentheses correctly. This was found in a real program which works in
> Dr. Racket in R5RS mode. You are not the only interpreter on Fedora 15
> which fails this, but anyhow.
>
> Testcase:
>
> (define bug
>  '(\())
>
> Expected results:
> No errors. bug variable should now be bound to the list (().
>
> Actual result:
> [torstein@Desktor Skrivebord]$ guile -l testcase.scm  --no-auto-compile
> ERROR: In procedure primitive-load:
> ERROR: In procedure scm_i_lreadparen:
> /home/torstein/Skrivebord/testcase.scm:3:12: end of file
>
> In the actual program the list was larger, '(\{ \[ \(), and then the error was:
> ERROR: In procedure primitive-load:
> ERROR: In procedure scm_read_expression:
> /home/torstein/Skrivebord/testcase.scm:2:27: unexpected ")"

Whether or not this is a bug, I cannot say, but I would like to point
out that neither the R5RS nor the R6RS consider '\( to be a valid symbol.

I shan't quote the grammars here, but you can confirm this for yourself.
http://www.schemers.org/Documents/Standards/R5RS/HTML/r5rs-Z-H-10.html#%_sec_7.1.2
http://www.r6rs.org/final/html/r6rs/r6rs-Z-H-7.html#node_sec_4.3.1

Specifically, in the R5RS, \ is not an initial (a letter or a special
initial), nor a peculiar identifier. In the R6RS, it is not an initial
(a constituent, a special initial, or an inline hex escape) nor a
peculiar identifier.

Handling additional identifiers is done in different ways in different
Scheme implementations. '|(| is pretty common, but in guile you would use
'#{(}#. If you want portable R5RS though, you should use
(string->symbol  "(")). If you choose R6RS, you can also use hex
escapes, in this case '\x28;

Also, apologies for the double send, but I forgot to include the mailing list.

Hope that helps,
Ian