Lutz Euler | 1 Jun 21:06 2012
Picon

Re: Expected failure of test "float.pure.lisp/RANGE-REDUCTION"

Nathan Froyd wrote,

> On Tue, May 29, 2012 at 2:32 PM, Lutz Euler <lutz.euler <at> freenet.de> wrote:
> > Maybe we should just make the expected test results dependent on whether
> > the platform is x86 or not. With a bit of multi-precision arithmetic I
> > can build a reducer modulo 2 pi with sufficient precision to replicate
> > the results of the libm range-reduction and thus test that.
> >
> > Just now I see that you enabled the test originally only for x86.
> >
> > So another solution would be just to make the test x86-only again?
> 
> I think marking the test as requiring x86-only weirdness by whatever
> means would be fine.

I have prepared two patches: The first marks the test skipped-on non-x86
and adds a comment as to why, and the second one adds another test,
skipped-on x86, that tests the libm range-reduction.

If no one objects, I will commit both in a few days.

Kind regards,

Lutz

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
(Continue reading)

Lutz Euler | 4 Jun 02:52 2012
Picon

Use different random numbers each time through the test suite.

Hi,

on 2012-05-01 we were talking on IRC about changing the test runner so
that it uses different random numbers each time (to "explore more of the
space", quoting Christophe), provided there is an easy way to repeat a
run with the same numbers.

I have built a preliminary something to this effect, using Nikodemus's
suggestion to generate a random state from a random integer seed, and
uploaded it to github:

https://github.com/leuler/sbcl/tree/random-random-tests

There are two commits, the first some cleanups, the second the new
functionality. I'd appreciate feedback especially on whether this does
what you'd like it to do and on the user interface.

It works with pure and impure tests. I have not yet looked into passing
the random seed/state to shell scripts called by the test runner, too,
but plan to do that.

To test it, just build a pure or impure test file with contents

(with-test (:name :random)
  (assert (zerop (random 2))))

and run it a few times. See the comment at the top of run-tests.sh for
how to pass a random seed into the script.

Regards,
(Continue reading)

Lutz Euler | 7 Jun 14:45 2012
Picon

Re: Expected failure of test "float.pure.lisp/RANGE-REDUCTION"

Hi,

I wrote:

> I have prepared two patches: The first marks the test skipped-on non-x86
> and adds a comment as to why, and the second one adds another test,
> skipped-on x86, that tests the libm range-reduction.

I have just committed these two. I am curious to learn if any
combinations of backend and operating system fail the second test
(the one added with commit 4de15256dd6387e5) as I tested it only
on x86-64 under Linux.

Greetings,

Lutz

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
Eric Marsden | 8 Jun 12:40 2012
Picon

GC invariant lost, gc-common.c:191

Hi,

I am running repeatedly into the following error on AMD64 when running
some intensive compilation, with SBCL 1.0.57.32-7535b05.

,----
| Final object pointer 0x5030053ce0, start 0x10099a8000, end 0x10099b0000
| fatal error encountered in SBCL pid 7080(tid 140737354012416):
| GC invariant lost, file "gc-common.c", line 191
| Welcome to LDB, a low-level debugger for the Lisp runtime environment.
| ldb> 
`----

--

-- 
Eric Marsden

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
Martin Cracauer | 9 Jun 01:07 2012

Re: GC invariant lost, gc-common.c:191

Eric Marsden wrote on Fri, Jun 08, 2012 at 12:40:45PM +0200: 
> Hi,
> 
> I am running repeatedly into the following error on AMD64 when running
> some intensive compilation, with SBCL 1.0.57.32-7535b05.
> 
> ,----
> | Final object pointer 0x5030053ce0, start 0x10099a8000, end 0x10099b0000
> | fatal error encountered in SBCL pid 7080(tid 140737354012416):
> | GC invariant lost, file "gc-common.c", line 191
> | Welcome to LDB, a low-level debugger for the Lisp runtime environment.
> | ldb> 
> `----

Can you re-run this after compiling the compiler with safety=3?

Martin
--

-- 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Martin Cracauer <cracauer <at> cons.org>   http://www.cons.org/cracauer/

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
Nikodemus Siivola | 9 Jun 12:36 2012
Picon

Fwd: [cmucl-imp] Unboxed float arguments

FYI, in case someone isn't reading cmucl-impl...

Cheers,

 -- nikodemus

---------- Forwarded message ----------
From: Helmut Eller <heller <at> common-lisp.net>
Date: 9 June 2012 10:46
Subject: [cmucl-imp] Unboxed float arguments
To: cmucl-imp <at> cons.org

Here is an idea for a calling convention that supports unboxed floats:

1. When DEFUN compiles a function that has a ftype declaration involving
floats, it creates a special entry-point that accepts unboxed arguments.

2. Named calls to such functions are compiled so that the arguments are
represented as indicated by the ftype declaration.

3. We also keep a linker table to connect those special entry-points to
call-sites.  When the function gets redefined, we go through the
existing call-sites and patch them so that the new definition is called.
If the type of the new definition is different than the one expected by
the call-site, we create an "adapter function" that converts the
representation to the new type or if the conversion is not possible
signals an error.

Attached is a prototype implementation to illustrate how this could be
done.  See the comments in the file for the main points.  A more
(Continue reading)

Lutz Euler | 9 Jun 23:34 2012
Picon

Error in tools-for-build/whitespacely-canonical-filenames

Hi,

since commit 9f4fead3c823ac5470bd7619d69d055c75c36e1f,
"In whitespace canonicalization, find a suitable make automatically",
I get the following error when running "sh make.sh":

tools-for-build/whitespacely-canonical-filenames: 26: Syntax error: ";" unexpected

This is on Ubuntu Linux, the shell is dash 0.5.5.1-7.2ubuntu1.
A possible fix is to change the semicolon at the end of line 26 to a
colon:

Old:
if   [ -n "$GNUFIND" ];        then ;

New:
if   [ -n "$GNUFIND" ];        then :

What is disconcerting for me here is that make.sh happily works on to
build SBCL. Shouldn't it check whether subcommands succeed and terminate
immediately if not? I only accidentally noticed the problem with
whitespacely-canonical-filenames when looking at the start of the build
output for other reasons.

Greetings,

Lutz

------------------------------------------------------------------------------
Live Security Virtual Conference
(Continue reading)

David Lichteblau | 10 Jun 00:23 2012

Re: Error in tools-for-build/whitespacely-canonical-filenames

Hi,

thank you for the bug report.  I hope I have managed to push it without
typos this time...

Quoting Lutz Euler (lutz.euler <at> freenet.de):
> What is disconcerting for me here is that make.sh happily works on to
> build SBCL. Shouldn't it check whether subcommands succeed and terminate
> immediately if not? I only accidentally noticed the problem with
> whitespacely-canonical-filenames when looking at the start of the build
> output for other reasons.

Well, whitespacely-canonical-filenames lacks set -e.  However, that alone
wouldn't help, because canonicalize-whitespace uses it in a pipe.

The easy way to check an exit code of a command in a pipeline is:
  ${PIPESTATUS[*]}
but that is bash-specific IIUC.

Would it be OK to use #!/bin/bash in our scripts?  My Solaris happens to
have /bin/bash, but that's probably a very new feature.

d.

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
(Continue reading)

Nikodemus Siivola | 10 Jun 11:08 2012
Picon

Re: Error in tools-for-build/whitespacely-canonical-filenames

On 10 June 2012 01:23, David Lichteblau <david <at> lichteblau.com> wrote:

> Would it be OK to use #!/bin/bash in our scripts?  My Solaris happens to
> have /bin/bash, but that's probably a very new feature.

I think having a build-dependency on bash would not be terribly
onerous -- so I'd be fine with it.

...but I realize I'm not among those negatively affected, so...

Cheers,

 -- Nikodemus

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Sbcl-devel mailing list
Sbcl-devel <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-devel
Nikodemus Siivola | 12 Jun 08:13 2012
Picon

Re: sb-pcl:fast-method's in the slime backtrace

On 29 May 2012 20:49, Attila Lendvai <attila.lendvai <at> gmail.com> wrote:
>> What's the pull-url?
>>
>> I'll try to get at least some of these in this month.
>
> sorry, i got used to darcsweb which is advertising that...
>
> git://dwim.hu/sbcl

Progress. I've merged one of my long-pending backtrace cleanups to
make way for this... Hopefully I can manage to move this a couple of
steps further over the next few days.

Cheers,

 -- nikodemus

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/

Gmane