Nikodemus Siivola | 1 Dec 12:46 2011
Picon

Re: BUG: LVAR-VALUE used on non-constant LVAR

On 1 December 2011 00:05, Eric Marsden <eric.marsden <at> free.fr> wrote:

> (lambda (p1 p3 p4)
>     (declare (type keyword p3))
>     (tree-equal p1 (cons 1 2) (the (member :test) p3) p4))

Thanks! Fixed in my pending tree.

Cheers,

 -- Nikodemus

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure 
contains a definitive record of customers, application performance, 
security threats, fraudulent activity, and more. Splunk takes this 
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
_______________________________________________
Sbcl-devel mailing list
Sbcl-devel <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-devel
Alastair Bridgewater | 1 Dec 15:09 2011
Picon

Re: BUG: illegal instruction on PowerPC

Confirmed, session transcript follows.

The interesting part of the disassembly is where execution basically
"falls off the end" of a block and into an LRA header (the transition
between DF8 and DFC).

This is on a PowerMac G4 running Linux.

I don't have the time to look into this seriously over the next few
days, but can produce compiler trace output if it would help.

-- Alastair Bridgewater

* (lambda (a) (declare (optimize (debug 2)) (type (member 2d0 1d0) a))
(truncate a))

#<FUNCTION (LAMBDA (A)) {510B28B5}>
* (sb-disassem:disassemble-code-component *)

; 510B28B0:       .ENTRY (LAMBDA (A))(A)                      ; (FUNCTION (#) *)
;      8CC:       3A7FFFC3         ADDI $CODE,$LIP,-61
;      8D0:       3A0F0040         ADDI $CSP,$CFP,64
;      8D4:       916F0010         STW $NARGS,16($CFP)
;      8D8:       930F000C         STW $A0,12($CFP)
;      8DC:       92CF0000         STW $OCFP,0($CFP)
;      8E0:       92EF0004         STW $LRA,4($CFP)

;;; [7] (TRUNCATE A)

;      8E4:       806F0010         LWZ $NL0,16($CFP)
(Continue reading)

Alastair Bridgewater | 1 Dec 15:19 2011
Picon

Re: BUG: illegal instruction on PowerPC

Did I say DF8 to DFC? I meant 9E8 to 9EC.

Additionally, this appears to be a regression since 1.0.28.

-- Alastair Bridgewater

On Thu, Dec 1, 2011 at 9:09 AM, Alastair Bridgewater
<alastair.bridgewater <at> gmail.com> wrote:
> Confirmed, session transcript follows.
>
> The interesting part of the disassembly is where execution basically
> "falls off the end" of a block and into an LRA header (the transition
> between DF8 and DFC).
>
> This is on a PowerMac G4 running Linux.
>
> I don't have the time to look into this seriously over the next few
> days, but can produce compiler trace output if it would help.
>
> -- Alastair Bridgewater

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure 
contains a definitive record of customers, application performance, 
security threats, fraudulent activity, and more. Splunk takes this 
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
Nikodemus Siivola | 1 Dec 17:17 2011
Picon

Re: BUG: illegal instruction on PowerPC

Here's a "portable" version of the test-case:

(funcall (compile nil
                  `(lambda (a)
                     (declare (type (member 1d0 2d0) a))
                     (block return-value-tag
                       (funcall
                         (the function
                           (catch 'debug-catch-tag
                             (return-from return-value-tag
                               (progn (truncate a)))))))))
                  2d0)

The bug's in the TRUNCATE transform when it sees *WILD-TYPE* as the result type.

Cheers,

 -- nikodemus

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure 
contains a definitive record of customers, application performance, 
security threats, fraudulent activity, and more. Splunk takes this 
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
Martin Cracauer | 2 Dec 18:57 2011

Re: freeze for 1.0.54

I thought this was fixed but I still have these spurious output
messages in 07:

(:REF
 #<SB-C::REF
   :%SOURCE-NAME SB-C::SEQ
   :LEAF #<SB-C::LAMBDA-VAR
           :%SOURCE-NAME SB-C::SEQ
           :TYPE #<SB-KERNEL:ARRAY-TYPE (SIMPLE-VECTOR 6)>
	   {102A179B43}>
   {102A17C0F3}>
 #<SB-C::LVAR 1 {102A17D863}>) 
(:REF
 #<SB-C::REF
   :%SOURCE-NAME SB-C::SEQ
   :LEAF #<SB-C::LAMBDA-VAR
           :%SOURCE-NAME SB-C::SEQ
           :TYPE #<SB-KERNEL:ARRAY-TYPE (SIMPLE-VECTOR 6)>
	   {102A217A33}>
   {102A21A023}>
 #<SB-C::LVAR 2 {102A21B793}>) 
(:REF
 #<SB-C::REF
   :%SOURCE-NAME SB-C::SEQ
   :LEAF #<SB-C::LAMBDA-VAR
           :%SOURCE-NAME SB-C::SEQ
           :TYPE #<SB-KERNEL:ARRAY-TYPE (SIMPLE-VECTOR 6)>
	   {102A08DDF3}>
   {102A090C33}>
 #<SB-C::LVAR 3 {102A092473}>) 
(Continue reading)

Nikodemus Siivola | 2 Dec 23:25 2011
Picon

Re: freeze for 1.0.54

On 2 December 2011 19:57, Martin Cracauer <cracauer <at> cons.org> wrote:

> I thought this was fixed but I still have these spurious output
> messages in 07:

Wow. This has been there since 1.0.48.21. Pushed a fix.

Cheers,

 -- nikodemus

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure 
contains a definitive record of customers, application performance, 
security threats, fraudulent activity, and more. Splunk takes this 
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
Attila Lendvai | 3 Dec 06:14 2011
Picon

Re: freeze for 1.0.54

>> I thought this was fixed but I still have these spurious output
>> messages in 07:
>
> Wow. This has been there since 1.0.48.21. Pushed a fix.

for your copy-pasting convenience, i have these against such accidents:

(font-lock-add-keywords
   'lisp-mode
   `(("[^-]\\(FIXME\\|TODO\\|KLUDGE\\|QUESTION\\|WARNING\\|#[+-]debug\\)"
1 'font-lock-todo-face t)
     ("[(:]\\(in-package\\|in-suite\\|not-yet-implemented\\|ignore-errors\\|read-from-string\\|eval\\|production-only.?\\|break[^)]?\\|break/inspect.?\\|break/print.?\\|print\\)[
	\n()]" 1 'font-lock-todo-face t)
     ("\\*debug-io\\*\\|\\*trace-output\\*" 0 'font-lock-todo-face t)))

--

-- 
 attila

Notice the erosion of your (digital) freedom, and do something about it!

PGP: 2FA1 A9DC 9C1E BA25 A59C  963F 5D5F 45C7 DFCD 0A39
OTR XMPP: 8647EEAC EA30FEEF E1B55146 573E52EE 21B1FF06

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure 
contains a definitive record of customers, application performance, 
security threats, fraudulent activity, and more. Splunk takes this 
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
_______________________________________________
(Continue reading)

Nikodemus Siivola | 5 Dec 10:48 2011
Picon

Re: BUG: illegal instruction on PowerPC

On 1 December 2011 18:17, Nikodemus Siivola <nikodemus <at> random-state.net> wrote:

> The bug's in the TRUNCATE transform when it sees *WILD-TYPE* as the result type.

Fix pushed.

Cheers,

 -- Nikodemus

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure 
contains a definitive record of customers, application performance, 
security threats, fraudulent activity, and more. Splunk takes this 
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
_______________________________________________
Sbcl-devel mailing list
Sbcl-devel <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-devel
Nikodemus Siivola | 5 Dec 10:49 2011
Picon

Re: BUG: incorrect results from FFLOOR (AMD64)

On 30 November 2011 16:00, Nikodemus Siivola <nikodemus <at> random-state.net> wrote:
> In the meanwhile,
>
>  https://github.com/nikodemus/SBCL/commit/4ecc79c0379b7454f7d8cf67054c2629643426e1
>
> has the fix if someone needs it in a hurry.

Pushed,

 -- nikodemus

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure 
contains a definitive record of customers, application performance, 
security threats, fraudulent activity, and more. Splunk takes this 
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
_______________________________________________
Sbcl-devel mailing list
Sbcl-devel <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-devel
Nikodemus Siivola | 5 Dec 10:48 2011
Picon

Re: BUG: LVAR-VALUE used on non-constant LVAR

On 1 December 2011 13:46, Nikodemus Siivola <nikodemus <at> random-state.net> wrote:
> On 1 December 2011 00:05, Eric Marsden <eric.marsden <at> free.fr> wrote:
>
>> (lambda (p1 p3 p4)
>>     (declare (type keyword p3))
>>     (tree-equal p1 (cons 1 2) (the (member :test) p3) p4))
>
> Thanks! Fixed in my pending tree.

Pushed,

 -- nikodemus

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure 
contains a definitive record of customers, application performance, 
security threats, fraudulent activity, and more. Splunk takes this 
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
_______________________________________________
Sbcl-devel mailing list
Sbcl-devel <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-devel

Gmane