Re: [Ecls-list] BUG: SEGMENTATION-VIOLATION in code with DEFTYPE unicode char
jgr == Juan Jose Garcia-Ripoll juanjose.garciarip...@googlemail.com writes: jgr Here is the problem: char in Linux is signed char, while in OS X it jgr seems to default to unsigned char. I have changed ECL so that BASE-CHAR jgr objects are unboxed using the explicit C type unsigned char to avoid such jgr ambiguities. ASA I get to a non-firewalled computer I will upload the fix. Thanks for the fix. Here is another bug which is possibly related (still on Linux/AMD64). , | ECL (Embeddable Common-Lisp) 11.1.1 (git:78442fa7bcb4ef486b704e16d0e7cefbd4bf7680) | Top level. | (lambda (a) (declare (type (eql b) a)) (eql b a)) | #bytecompiled-closure #bytecompiled-function 03cc6e10 | (compile nil *) | ;;; OPTIMIZE levels: Safety=2, Space=0, Speed=3, Debug=0 | ;;; | ;;; End of Pass 1. | #compiled-function 03fc6980 | NIL | NIL | (funcall * b) | Condition of type: TYPE-ERROR | b is not of type (EQL b). ` -- Eric Marsden -- 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 ___ Ecls-list mailing list Ecls-list@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ecls-list
Re: [Ecls-list] BUG: SEGMENTATION-VIOLATION in code with DEFTYPE unicode char
Eric Marsden eric.mars...@free.fr writes: jgr == Juan Jose Garcia-Ripoll juanjose.garciarip...@googlemail.com writes: jgr Here is the problem: char in Linux is signed char, while in OS X it jgr seems to default to unsigned char. I have changed ECL so that BASE-CHAR jgr objects are unboxed using the explicit C type unsigned char to avoid such jgr ambiguities. ASA I get to a non-firewalled computer I will upload the fix. Thanks for the fix. Here is another bug which is possibly related (still on Linux/AMD64). , | ECL (Embeddable Common-Lisp) 11.1.1 (git:78442fa7bcb4ef486b704e16d0e7cefbd4bf7680) | Top level. | (lambda (a) (declare (type (eql b) a)) (eql b a)) | #bytecompiled-closure #bytecompiled-function 03cc6e10 | (compile nil *) | ;;; OPTIMIZE levels: Safety=2, Space=0, Speed=3, Debug=0 | ;;; | ;;; End of Pass 1. | #compiled-function 03fc6980 | NIL | NIL | (funcall * b) | Condition of type: TYPE-ERROR | b is not of type (EQL b). ` I don't see how that's a bug, b isn't required to be EQL to b. It can be, especially when coalesced by the compiler, but the reader usually constructs a new string each time. -- With best regards, Stas. -- 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 ___ Ecls-list mailing list Ecls-list@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ecls-list
Re: [Ecls-list] BUG: SEGMENTATION-VIOLATION in code with DEFTYPE unicode char
On Tue, Nov 22, 2011 at 10:47 AM, Stas Boukarev stass...@gmail.com wrote: I don't see how that's a bug, b isn't required to be EQL to b. It can be, especially when coalesced by the compiler, but the reader usually constructs a new string each time. I agree. Compare that with (lambda (a) (declare (type (eql #1=foo) a)) (eql '#1# a)) #bytecompiled-closure #bytecompiled-function 000103cbfa00 (compile 'foo *) ;;; OPTIMIZE levels: Safety=2, Space=0, Speed=3, Debug=0 ;;; ;;; End of Pass 1. FOO NIL NIL (foo 'foo) T -- Instituto de Física Fundamental, CSIC c/ Serrano, 113b, Madrid 28006 (Spain) http://juanjose.garciaripoll.googlepages.com -- 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___ Ecls-list mailing list Ecls-list@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ecls-list
Re: [Ecls-list] BUG: SEGMENTATION-VIOLATION in code with DEFTYPE unicode char
jgr == Juan Jose Garcia-Ripoll juanjose.garciarip...@googlemail.com writes: jgr I agree. Compare that with Indeed, insufficient thinking on my part, sorry. -- Eric Marsden -- 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 ___ Ecls-list mailing list Ecls-list@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ecls-list
Re: [Ecls-list] BUG: SEGMENTATION-VIOLATION in code with DEFTYPE unicode char
On Sun, Nov 13, 2011 at 7:18 PM, Eric Marsden eric.mars...@free.fr wrote:
,
| (disassemble '(lambda (p1 p2) (declare (optimize (safety 0)) (type
(member #\U00FE) p2)) (equalp p1 p2)))
|
| ;;; Compiling (DEFUN C::GAZONK ...).
| ;;; Emitting code for C::GAZONK.
| /*function definition for GAZONK*/
| /*optimize speed 3, debug 0, space 0, safety 0 */
| static cl_object L1c__gazonk(cl_object V1, cl_object V2)
| { VT2 VLEX2 CLSR2 STCK2
| const cl_env_ptr cl_env_copy = ecl_process_env();
| cl_object value0;
| char V3;
| V3= CHAR_CODE(V2);
Here is the problem: char in Linux is signed char, while in OS X it
seems to default to unsigned char. I have changed ECL so that BASE-CHAR
objects are unboxed using the explicit C type unsigned char to avoid such
ambiguities. ASA I get to a non-firewalled computer I will upload the fix.
Juanjo
--
Instituto de Física Fundamental, CSIC
c/ Serrano, 113b, Madrid 28006 (Spain)
http://juanjose.garciaripoll.googlepages.com
--
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___
Ecls-list mailing list
Ecls-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ecls-list
Re: [Ecls-list] BUG: SEGMENTATION-VIOLATION in code with DEFTYPE unicode char
jgr == Juan Jose Garcia-Ripoll juanjose.garciarip...@googlemail.com
writes:
jgr This will give us a clue about the quality of the generated code -- I am
jgr asking because in my OS X it looks just fine.
Here is the result of attaching gdb.
,
| Program received signal SIGSEGV, Segmentation fault.
| [Switching to Thread 0x7f256ac2c700 (LWP 27462)]
| ecl_char_upcase (code=-2) at /usr/local/src/ecl/src/c/char_ctype.d:171
| 171 if (c[0] == 1) {
| (gdb) print c[0]
| Cannot access memory at address 0x3f8
| (gdb) disassemble
| Dump of assembler code for function ecl_char_upcase:
|0x7f256a61ccc0 +0: push %rbx
|0x7f256a61ccc1 +1: mov%edi,%ebx
|0x7f256a61ccc3 +3: callq 0x7f256a53f720 ucd_char_data@plt
| = 0x7f256a61ccc8 +8: cmpb $0x1,(%rax)
|0x7f256a61cccb +11: jne0x7f256a61cce3 ecl_char_upcase+35
|0x7f256a61cccd +13: movzbl 0x2(%rax),%edx
|0x7f256a61ccd1 +17: movzbl 0x1(%rax),%ebx
|0x7f256a61ccd5 +21: movzbl 0x3(%rax),%eax
|0x7f256a61ccd9 +25: shl$0x8,%edx
|0x7f256a61ccdc +28: add%edx,%ebx
|0x7f256a61ccde +30: shl$0x10,%eax
|0x7f256a61cce1 +33: add%eax,%ebx
|0x7f256a61cce3 +35: mov%ebx,%eax
|0x7f256a61cce5 +37: pop%rbx
|0x7f256a61cce6 +38: retq
| End of assembler dump.
`
--
Eric Marsden
--
RSA(R) Conference 2012
Save $700 by Nov 18
Register now
http://p.sf.net/sfu/rsa-sfdev2dev1
___
Ecls-list mailing list
Ecls-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ecls-list
Re: [Ecls-list] BUG: SEGMENTATION-VIOLATION in code with DEFTYPE unicode char
On Sat, Nov 12, 2011 at 12:47 PM, Eric Marsden eric.mars...@free.fr wrote: The following transcript illustrates a bug (on Linux/AMD64) when compiling code with low safety and type declarations involving unicode characters. A few things would be helpful - Configuration flags for ECL - Value of *features* - Output of DISASSEMBLE Juanjo -- Instituto de Física Fundamental, CSIC c/ Serrano, 113b, Madrid 28006 (Spain) http://juanjose.garciaripoll.googlepages.com -- RSA(R) Conference 2012 Save $700 by Nov 18 Register now http://p.sf.net/sfu/rsa-sfdev2dev1___ Ecls-list mailing list Ecls-list@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ecls-list
Re: [Ecls-list] BUG: SEGMENTATION-VIOLATION in code with DEFTYPE unicode char
On Sun, Nov 13, 2011 at 7:06 PM, Eric Marsden eric.mars...@free.fr wrote: The bytecode disassembly of my function is below. However, the function works fine when byte-compiled, only segfaulting when using the C compiler. I haven't managed to obtain disassembly when compiled via C. Sorry, what I meant was the C disassembly. You get it with (require :cmp) (disassemble '(lambda (...) ..)) This will give us a clue about the quality of the generated code -- I am asking because in my OS X it looks just fine. Juanjo -- Instituto de Física Fundamental, CSIC c/ Serrano, 113b, Madrid 28006 (Spain) http://juanjose.garciaripoll.googlepages.com -- RSA(R) Conference 2012 Save $700 by Nov 18 Register now http://p.sf.net/sfu/rsa-sfdev2dev1___ Ecls-list mailing list Ecls-list@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ecls-list
