caml-list - the Caml user's mailing list
 help / color / mirror / Atom feed
* [Caml-list] Cash broken?
@ 2004-05-27 19:38 John Goerzen
  2004-05-28 10:26 ` Bruno.Verlyck
  0 siblings, 1 reply; 4+ messages in thread
From: John Goerzen @ 2004-05-27 19:38 UTC (permalink / raw)
  To: caml-list

I tried this example straight from the docs:

$ cashtop
        Objective Caml version 3.07+2

        Camlp4 Parsing version 3.07+2

# open Cash;;
# run_with_strings (fun () -> exec_path "ls" []);;
Exception: Sys_error "Bad file descriptor".
# Exception: Sys_error "Bad file descriptor".

I can't make any of the run_with_* that I've tried work, whether in my
programs or with cashtop.

Ideas?  (Cash 0.20)

-- John

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [Caml-list] Cash broken?
  2004-05-27 19:38 [Caml-list] Cash broken? John Goerzen
@ 2004-05-28 10:26 ` Bruno.Verlyck
  2004-05-28 13:13   ` John Goerzen
  0 siblings, 1 reply; 4+ messages in thread
From: Bruno.Verlyck @ 2004-05-28 10:26 UTC (permalink / raw)
  To: jgoerzen; +Cc: caml-list

[-- Attachment #1: Type: text/plain, Size: 1097 bytes --]

   Date: Thu, 27 May 2004 14:38:23 -0500
   From: John Goerzen <jgoerzen@complete.org>

Hi John,
   I tried this example straight from the docs:
    $ cashtop
	    Objective Caml version 3.07+2
	    Camlp4 Parsing version 3.07+2
    # open Cash;;
    # run_with_strings (fun () -> exec_path "ls" []);;
    Exception: Sys_error "Bad file descriptor".
    # Exception: Sys_error "Bad file descriptor".
   I can't make any of the run_with_* that I've tried work, whether in
   my programs or with cashtop.

   Ideas?  (Cash 0.20)
Not a lot with so little info (of course it works here).  Would you
mind sending me more OS details ?  A backtrace would be very useful,
and/or the result of trace/truss/strace -f ?

That said, what comes to mind is that the behavior of close_{in,out}
has changed in Pervasives over the last versions of Caml.  I wrote
some (too many) lines of code to make Cash.close_{in,out} work anyway.
I see in the CVS log that there was a change related to that, some
time after 3.07 (see the attached patch).  You could try it, or even
try the entire CVS version of Cash.

Bruno.

[-- Attachment #2: Type: text/plain, Size: 1308 bytes --]

Index: io_3_2.ml
===================================================================
RCS file: /net/pauillac/caml/repository/cash/io_3_2.ml,v
retrieving revision 1.47
retrieving revision 1.48
diff -c -a -r1.47 -r1.48
*** io_3_2.ml	2002/08/05 17:07:37	1.47
--- io_3_2.ml	2002/12/05 17:37:52	1.48
***************
*** 116,126 ****
        Some fd ->
          let res = low_close_fd fd in  (* low_close_fd will tell it. *)
          do {
!           close_any any_chan; (* Never raise exception in the current implementation. *)
            res
          } ]
    in
    (fun ichan -> close (anychan_of_ichan ichan),
     fun ochan -> do { flush ochan; close (anychan_of_ochan ochan) })
  ;
  
--- 116,129 ----
        Some fd ->
          let res = low_close_fd fd in  (* low_close_fd will tell it. *)
          do {
!           (* Used to raise no exception, but this varies between implementations... *)
!           try close_any any_chan with [ Sys_error "Bad file descriptor" -> () ];
            res
          } ]
    in
    (fun ichan -> close (anychan_of_ichan ichan),
+    (* Current religion says: flush do nothing on a closed channel. *)
+    (* Current implication is: if an unflushed channel's fd is closed, it's an error. *)
     fun ochan -> do { flush ochan; close (anychan_of_ochan ochan) })
  ;

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [Caml-list] Cash broken?
  2004-05-28 10:26 ` Bruno.Verlyck
@ 2004-05-28 13:13   ` John Goerzen
  2004-05-28 13:52     ` Bruno.Verlyck
  0 siblings, 1 reply; 4+ messages in thread
From: John Goerzen @ 2004-05-28 13:13 UTC (permalink / raw)
  To: Bruno.Verlyck; +Cc: caml-list

[-- Attachment #1: Type: text/plain, Size: 2699 bytes --]

On Fri, May 28, 2004 at 12:26:15PM +0200, Bruno.Verlyck@inria.fr wrote:
>     # open Cash;;
>     # run_with_strings (fun () -> exec_path "ls" []);;
>     Exception: Sys_error "Bad file descriptor".
>     # Exception: Sys_error "Bad file descriptor".
>    I can't make any of the run_with_* that I've tried work, whether in
>    my programs or with cashtop.
> 
>    Ideas?  (Cash 0.20)
> Not a lot with so little info (of course it works here).  Would you
> mind sending me more OS details ?  A backtrace would be very useful,
> and/or the result of trace/truss/strace -f ?

This is on Debian sid, Linux 2.6.6.  I couldn't figure out how to get a
backtrace.  However, from the forked child, I see:

[pid  9461] close(4)                    = 0
[pid  9461] close(4)                    = -1 EBADF (Bad file descriptor)

(Full strace attached)

Will try the patch.  How does the CVS version rank in terms of
stability?

> That said, what comes to mind is that the behavior of close_{in,out}
> has changed in Pervasives over the last versions of Caml.  I wrote
> some (too many) lines of code to make Cash.close_{in,out} work anyway.
> I see in the CVS log that there was a change related to that, some
> time after 3.07 (see the attached patch).  You could try it, or even
> try the entire CVS version of Cash.
> 
> Bruno.

> Index: io_3_2.ml
> ===================================================================
> RCS file: /net/pauillac/caml/repository/cash/io_3_2.ml,v
> retrieving revision 1.47
> retrieving revision 1.48
> diff -c -a -r1.47 -r1.48
> *** io_3_2.ml	2002/08/05 17:07:37	1.47
> --- io_3_2.ml	2002/12/05 17:37:52	1.48
> ***************
> *** 116,126 ****
>         Some fd ->
>           let res = low_close_fd fd in  (* low_close_fd will tell it. *)
>           do {
> !           close_any any_chan; (* Never raise exception in the current implementation. *)
>             res
>           } ]
>     in
>     (fun ichan -> close (anychan_of_ichan ichan),
>      fun ochan -> do { flush ochan; close (anychan_of_ochan ochan) })
>   ;
>   
> --- 116,129 ----
>         Some fd ->
>           let res = low_close_fd fd in  (* low_close_fd will tell it. *)
>           do {
> !           (* Used to raise no exception, but this varies between implementations... *)
> !           try close_any any_chan with [ Sys_error "Bad file descriptor" -> () ];
>             res
>           } ]
>     in
>     (fun ichan -> close (anychan_of_ichan ichan),
> +    (* Current religion says: flush do nothing on a closed channel. *)
> +    (* Current implication is: if an unflushed channel's fd is closed, it's an error. *)
>      fun ochan -> do { flush ochan; close (anychan_of_ochan ochan) })
>   ;

[-- Attachment #2: foo.gz --]
[-- Type: application/octet-stream, Size: 4784 bytes --]

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [Caml-list] Cash broken?
  2004-05-28 13:13   ` John Goerzen
@ 2004-05-28 13:52     ` Bruno.Verlyck
  0 siblings, 0 replies; 4+ messages in thread
From: Bruno.Verlyck @ 2004-05-28 13:52 UTC (permalink / raw)
  To: jgoerzen; +Cc: caml-list

   Date: Fri, 28 May 2004 08:13:39 -0500
   From: John Goerzen <jgoerzen@complete.org>

   On Fri, May 28, 2004 at 12:26:15PM +0200, Bruno.Verlyck@inria.fr wrote:
   >     # open Cash;;
   >     # run_with_strings (fun () -> exec_path "ls" []);;
   >     Exception: Sys_error "Bad file descriptor".
   >     # Exception: Sys_error "Bad file descriptor".
   
   >    Ideas?  (Cash 0.20)
   > Would you mind sending me more OS details ?  A backtrace would be
   > very useful, and/or the result of trace/truss/strace -f ?
   This is on Debian sid, Linux 2.6.6.  I couldn't figure out how to
   get a backtrace.  However, from the forked child, I see:

   [pid  9461] close(4)			= 0
   [pid  9461] close(4)			= -1 EBADF (Bad file descriptor)

   (Full strace attached)
Thank you.  It confirms my first guess.

   Will try the patch.  How does the CVS version rank in terms of
   stability?
Pretty good :-).  The only risk is that I converted all the sources to
use Cstr/String13, and I have to admit I introduced some translation
errors.  I fixed those I found, but it could be that there's still
`just one more bug'.  I would be very interested if you (or anyone
else) find it out, or tell me everything you tried is OK.

Thanks anyway,
Bruno.

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners


^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2004-05-28 13:52 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2004-05-27 19:38 [Caml-list] Cash broken? John Goerzen
2004-05-28 10:26 ` Bruno.Verlyck
2004-05-28 13:13   ` John Goerzen
2004-05-28 13:52     ` Bruno.Verlyck

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).