That's because of avatars. It's wise to block them in one manner or another for that reason.bob wrote:The problem in CCC is that there are so many posts, and after just a few replies, everything shows up aligned on the right half of the page, which makes it easy to overlook something when it get lost in all the other posts. This is even in the wrong topic in fact...
Rybka 1.0 vs. Strelka
Moderator: Ras
-
bnemias
- Posts: 373
- Joined: Thu Aug 14, 2008 3:21 am
- Location: Albuquerque, NM
Re: Wanted: some opposition to the provided evidence
-
bob
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: Wanted: some opposition to the provided evidence
This just makes the source harder to read, just as if you went in with an editor and changed every variable name to something like a1111, a1112, a1113, etc...Sven Schüle wrote:Just to get the wordings correct ...Terry McCracken wrote:If Christophe or any other commercial programer wanted the source code of Rybka 3 they'd disassemble it. It's not that hard.
1) "Disassembling" (as also used by some others in this thread) only refers to converting the binary program into assembler instructions, like with the MSVC(++) command "dumpbin /disasm". What you mean is either "decompiling" or "reverse engineering".
2) You cannot obtain "the source code" of a binary via reverse engineering if the binary comes without a symbol table that usually contains (some of!) the identifiers the original author has chosen, and without further input. Since I only looked at the free R1.0beta exe I can only confirm absence of the symbol table for that version. But you can be 100% sure that the commercial versions have the symtable stripped off, too.
But that requires that one believe that these two lines are different:
3) So the reverse engineering is an ambiguous activity (i.e. there are many possible results being equivalent since they can be mapped to the same binary) which is in reality not fully automated, even if there might be some assisting tools for it. The reverse engineer has to assign identifiers to functions and (global/static) variables which are anonymous so far. And, what I personally consider to be the hardest job, he also has to deal quite a lot with local variables living either on the stack or in registers, and having no address that could simply be translated into an identifier.
if (a1111 < a1110)
a1112 = a12000 * a120001;
if (rook_square < pawn_square)
score = rook_behind_passed_pawn * passed_pawn_rank;
That would delight programmers wanting to copy assignments in programming courses if all they need to do is change variable names. Those of us with more experience might well deduce that the above are identical, assuming at any instant in time all the variable values are the same, or they could be completely different if all the values are different. It can be done quite easily. There are even freeware programs that will tell you if the above are equivalent, code designed to "half-compile" sources to get rid of comments and variable name hangens to compare the semantics as opposed to syntax.
So I propose to be very careful with a statement that describes reverse engineering of a bare executable as being "not that hard".
Just as additional info: the R1.0beta exe is about 2.4MB large (32 bit version). Strelka 2.0 32bit exe has about 180kB IIRC. Although the huge difference probably comes from the fact that Strelka 2.0 does not use some huge table that R1.0beta had used, this is still a lot, so IMO it _is_ hard work.
This remains true even with the help of available open source where you can borrow some function names from.
Just my 2 cents!
Sven
-
bob
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: Wanted: some opposition to the provided evidence
And toga, and Le Petite, and Voyager, and Bionic Impact, and who knows how many others?GenoM wrote:Wise words. Can I see Strelka as a legal engine now after all these discussions?chrisw wrote:
Sounds like witchhunting method. See if she floats. If she does, burn her for witchcraft, if she sinks, she drowns anyway.
See the source, if innocent, you get to see the source, if guilty you get to see the source. Commercial programmer loses either way, like the witch.
The onus of proof remains on the accusers. Period.
-
Marc Lacrosse
- Posts: 511
- Joined: Wed Mar 08, 2006 10:05 pm
Re: Wanted: some opposition to the provided evidence
Hmmm.
Vas just issued an interesting comment :
1. Strelka contains rybka code AND some other code
2. Rybka is and was not based on Fruit's code (there are similarities for commonly used ideas and algorithms but no Fruit code was used)
3. Maybe those parts of Strelka that are not made from disassembled rybka code do directly include code from fruit but this is not Vas' problem.
This is how I understand a post by Vas on his forum today (see exact copy at end.)
This explanation seems reasonable ...
From now on, only Ossipov (or someone having access to the real rybka code) will be able to tell the truth ...
With this statement it is in my eyes not interesting anymore to add demonstrations of similarities between strelka and Fruit, as long as at the same time there is no proof that the exposed code does not come from that part of strelka that has not been copied from dissassembled rybka ...
Marc
PS : Vas ' post at rybkachess today :
This discussion looks like a complete mess, I see wild statements like:
> --I noted earlier that the history initialization was different. Strelka and Rybka 1.0 are the same here, using memset(History, 256, 12 * 64 * sizeof(int)); But what I didn't notice is, 256 is out of range for an unsigned char. So that code is the same as Fruit's memset(History, 0, 12 * 64 * sizeof(int)); Why would Rybka 1.0 have 256? Is it the optimizer gone wild, or a deliberate attempt at obfuscation?
Maybe somebody can summarize the points.
Generally, code theft is easy to show - just show the two sections of identical code, side-by-side. There isn't much to debate in such cases. Rybka is of course original (with some accepted exceptions like bit scans & bit counters, etc). Strelka contains Rybka code. Whether Strelka also contains Fruit code, I don't know and don't really care.
Re. algorithms and structures - sure, you'll get a lot of similarity between programs. Take something like "modern move ordering" - by this, I mean hash move, non-losing captures (by MVV/LVA), killers, non-captures, and losing captures. The chance of two programmers settling on this independently is more or less zero.
Vas
Vas just issued an interesting comment :
1. Strelka contains rybka code AND some other code
2. Rybka is and was not based on Fruit's code (there are similarities for commonly used ideas and algorithms but no Fruit code was used)
3. Maybe those parts of Strelka that are not made from disassembled rybka code do directly include code from fruit but this is not Vas' problem.
This is how I understand a post by Vas on his forum today (see exact copy at end.)
This explanation seems reasonable ...
From now on, only Ossipov (or someone having access to the real rybka code) will be able to tell the truth ...
With this statement it is in my eyes not interesting anymore to add demonstrations of similarities between strelka and Fruit, as long as at the same time there is no proof that the exposed code does not come from that part of strelka that has not been copied from dissassembled rybka ...
Marc
PS : Vas ' post at rybkachess today :
This discussion looks like a complete mess, I see wild statements like:
> --I noted earlier that the history initialization was different. Strelka and Rybka 1.0 are the same here, using memset(History, 256, 12 * 64 * sizeof(int)); But what I didn't notice is, 256 is out of range for an unsigned char. So that code is the same as Fruit's memset(History, 0, 12 * 64 * sizeof(int)); Why would Rybka 1.0 have 256? Is it the optimizer gone wild, or a deliberate attempt at obfuscation?
Maybe somebody can summarize the points.
Generally, code theft is easy to show - just show the two sections of identical code, side-by-side. There isn't much to debate in such cases. Rybka is of course original (with some accepted exceptions like bit scans & bit counters, etc). Strelka contains Rybka code. Whether Strelka also contains Fruit code, I don't know and don't really care.
Re. algorithms and structures - sure, you'll get a lot of similarity between programs. Take something like "modern move ordering" - by this, I mean hash move, non-losing captures (by MVV/LVA), killers, non-captures, and losing captures. The chance of two programmers settling on this independently is more or less zero.
Vas
-
Sven
- Posts: 4052
- Joined: Thu May 15, 2008 9:57 pm
- Location: Berlin, Germany
- Full name: Sven Schüle
Re: Wanted: some opposition to the provided evidence
Maybe my post could be misunderstood such that I would focus on choice of identifiers only. In fact, my point was just to show that reverse engineering is not an easy task in general, as opposed to what Terry wrote.bob wrote:This just makes the source harder to read, just as if you went in with an editor and changed every variable name to something like a1111, a1112, a1113, etc...
[...]
But that requires that one believe that these two lines are different:
I have tried one of the most common free reverse engineering tools. I omit the name here to avoid advertising; maybe Rick Fadden used the same for his analysis that he posted here a few months ago. It is a great piece of software, but I think you still must be an assembler expert to work efficiently with it (I do not claim to be one), and even if you are, for me it looks as if you still have to work hard to get something that can be called a C source out of your binary, and that C source still compiles to the same original binary.
Of course you can create a different C source that has some common parts with the original binary but contains also a partial copy of another C source, where both worlds are put together somehow. Still hard work, I think.
Sven
-
bob
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: Wanted: some opposition to the provided evidence
Did you see Z's post doing just that???chrisw wrote:From Rybka forum, Vas states:bob wrote:The only possible "innocent explanation" would be "OK, I started by modifying fruit and I have modified the important parts, but didn't realize that the other parts would represent a problem". If you take any two chess programs on the planet, and compare the source, you are not going to find even a hundred lines of code that match exactly, much less a thousand or more. Unless they have a common ancestry, that is.chrisw wrote:What if there's a completely innocent explanation?kranium wrote:No, not everything is speculative.Enir wrote:Maybe. Reverse engineering Rybka 3 will prove whether or not it contains GPL code. Everything else is just speculative and damaging.bnemias wrote:If you need examples, Verizon sued by BusyBox developers. Redhat v SCO, Sun v Microsoft. These occur all the time because major players push software that isn't "squeaky clean." So it's hard to use Chessbase's intelligence as proof of anything concerning Rybka 3.tiger wrote:I think I have first hand experience on this and I know that you are wrong.chrisw wrote:[snip]
No commercial company with half a brain will publish software claiming total rights over it if not true. That's the overpowering evidence one that R3 is clean. Chessbase won't licence it unless absolutely certain it's clean. That's overpowering evidence two. Say anything else and you play with fire. imo.
Enrique
There's a mountain of evidence that strleka 2.0 has hundreds and hundreds of lines of code that are identical or equivilent to code found in fruit 2.1. (this comparison can be made even by a casual user...the two sets of source code are freely available). a detailed comparison was presented here:
http://64.68.157.89/forum/viewtopic.php ... ht=strelka
There is also evidence that strelka 2.0 (both source code and binary) is identical to rybka 1.0 beta.
In addition Vas has confirmed that Strelka is indeed his.
? I think the conclusion is not difficult to ascertain...
rybka 1.0 beta is a derivative of fruit 2.1
i volunteer to stir the shit....
Maybe somebody can summarize the points.
Generally, code theft is easy to show - just show the two sections of identical code, side-by-side. There isn't much to debate in such cases. Rybka is of course original (with some accepted exceptions like bit scans & bit counters, etc). Strelka contains Rybka code. Whether Strelka also contains Fruit code, I don't know and don't really care.
If you want to show a problem with Rybka, put it, Rybka side by side with Fruit.
-
bob
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: Wanted: some opposition to the provided evidence
The only fly in this ointment is that Zach has done exactly that between Fruit and Rybka 1. His post is in this forum near the top (and is pretty short thread so it is easy to find).
-
Terry McCracken
- Posts: 16465
- Joined: Wed Aug 01, 2007 4:16 am
- Location: Canada
Re: Rybka 1.0 vs. Strelka
Maybe not Bob...it gets lost fairly quickly.Zach Wegner wrote:The topic came up in another thread about the issue of comparing Strelka to Fruit vs. comparing Rybka 1.0 to Fruit. I am doing some disassembling of Rybka 1.0, so I will post the assembly that is equivalent to the last piece of code I had in Fruit. Just to let you know, this is by no means complete. I have filled in most of the function names and some of the variable names, but there is a lot left to do. This is just the result of a few day's work, more will come later.
The interesting thing that I found is that for this piece of code, Rybka 1.0 is more similar to Fruit than Strelka. Here are the differences:Code: Select all
.text:0040BCA0 start_search proc near ; CODE XREF: start_go+35F p .text:0040BCA0 .text:0040BCA0 var_44C = dword ptr -44Ch .text:0040BCA0 var_440 = dword ptr -440h .text:0040BCA0 undo = dword ptr -40h .text:0040BCA0 var_8 = dword ptr -8 .text:0040BCA0 num_moves = dword ptr -4 .text:0040BCA0 .text:0040BCA0 push ebp .text:0040BCA1 mov ebp, esp .text:0040BCA3 sub esp, 440h .text:0040BCA9 push ebx .text:0040BCAA push esi .text:0040BCAB push edi .text:0040BCAC push offset dword_669C60 .text:0040BCB1 mov dword_667A0C, -500 .text:0040BCBB mov dword_667A10, 500 .text:0040BCC5 call evaluate .text:0040BCCA add esp, 4 .text:0040BCCD test al, al .text:0040BCCF jz short loc_40BD25 .text:0040BCD1 mov ecx, dword_669C6C .text:0040BCD7 mov eax, dword_669C68 .text:0040BCDC not ecx .text:0040BCDE push ecx .text:0040BCDF not eax .text:0040BCE1 push eax .text:0040BCE2 push offset list_root .text:0040BCE7 call gen_evasions .text:0040BCEC mov ecx, list_root .text:0040BCF2 add esp, 12 .text:0040BCF5 xor eax, eax .text:0040BCF7 test ecx, ecx .text:0040BCF9 jz short loc_40BD15 .text:0040BCFB jmp short loc_40BD00 .text:0040BCFB ; --------------------------------------------------------------------------- .text:0040BCFD align 10h .text:0040BD00 .text:0040BD00 loc_40BD00: ; CODE XREF: start_search+5B j .text:0040BD00 ; start_search+73 j .text:0040BD00 mov list_root[eax*4], ecx .text:0040BD07 mov ecx, dword_66BC78[eax*8] .text:0040BD0E add eax, 1 .text:0040BD11 test ecx, ecx .text:0040BD13 jnz short loc_40BD00 .text:0040BD15 .text:0040BD15 loc_40BD15: ; CODE XREF: start_search+59 j .text:0040BD15 mov list_root[eax*4], 0 .text:0040BD20 jmp loc_40BDC2 .text:0040BD25 ; --------------------------------------------------------------------------- .text:0040BD25 .text:0040BD25 loc_40BD25: ; CODE XREF: start_search+2F j .text:0040BD25 mov edx, dword_667B9C .text:0040BD2B mov eax, dword_667B98 .text:0040BD30 mov ecx, dword_667B94 .text:0040BD36 push edx .text:0040BD37 mov edx, dword_667B90 .text:0040BD3D push eax .text:0040BD3E push ecx .text:0040BD3F push edx .text:0040BD40 push offset dword_66C070 .text:0040BD45 mov eax, offset list_root .text:0040BD4A call gen_captures .text:0040BD4F xor esi, esi .text:0040BD51 add esp, 20 .text:0040BD54 cmp list_root, esi .text:0040BD5A jz short loc_40BD6D .text:0040BD5C lea esp, [esp+0] .text:0040BD60 .text:0040BD60 loc_40BD60: ; CODE XREF: start_search+CB j .text:0040BD60 add esi, 1 .text:0040BD63 cmp list_root[esi*4], 0 .text:0040BD6B jnz short loc_40BD60 .text:0040BD6D .text:0040BD6D loc_40BD6D: ; CODE XREF: start_search+BA j .text:0040BD6D mov eax, dword_669C6C .text:0040BD72 mov ecx, dword_669C68 .text:0040BD78 push -1 .text:0040BD7A push -1 .text:0040BD7C push -1 .text:0040BD7E push eax .text:0040BD7F push ecx .text:0040BD80 lea edx, [ebp+var_440] .text:0040BD86 push edx .text:0040BD87 call gen_quiet_moves .text:0040BD8C mov eax, [ebp+var_440] .text:0040BD92 add esp, 24 .text:0040BD95 xor ecx, ecx .text:0040BD97 test eax, eax .text:0040BD99 jz short loc_40BDB5 .text:0040BD9B lea edx, ds:66BC70h[esi*4] .text:0040BDA2 .text:0040BDA2 loc_40BDA2: ; CODE XREF: start_search+113 j .text:0040BDA2 add ecx, 1 .text:0040BDA5 mov [edx], eax .text:0040BDA7 mov eax, [ebp+ecx*4+var_440] .text:0040BDAE add edx, 4 .text:0040BDB1 test eax, eax .text:0040BDB3 jnz short loc_40BDA2 .text:0040BDB5 .text:0040BDB5 loc_40BDB5: ; CODE XREF: start_search+F9 j .text:0040BDB5 add ecx, esi .text:0040BDB7 mov list_root[ecx*4], 0 .text:0040BDC2 .text:0040BDC2 loc_40BDC2: ; CODE XREF: start_search+80 j .text:0040BDC2 mov esi, list_root .text:0040BDC8 xor edi, edi .text:0040BDCA test esi, esi .text:0040BDCC mov [ebp+num_moves], edi .text:0040BDCF jz short loc_40BE4C .text:0040BDD1 mov [ebp+num_moves], offset list_root .text:0040BDD8 .text:0040BDD8 loc_40BDD8: ; CODE XREF: start_search+1A7 j .text:0040BDD8 lea eax, [ebp+undo] .text:0040BDDB push eax .text:0040BDDC push esi .text:0040BDDD call make_move .text:0040BDE2 push offset pos_info_entry .text:0040BDE7 call evaluate .text:0040BDEC mov ecx, dword_667C20 .text:0040BDF2 lea edx, ds:0[ecx*8] .text:0040BDF9 mov ecx, offset dword_667BF8 .text:0040BDFE sub ecx, edx .text:0040BE00 mov eax, [ecx] .text:0040BE02 mov ecx, [ecx+4] .text:0040BE05 and eax, pos_info_entry .text:0040BE0B and ecx, dword_669C84 .text:0040BE11 add esp, 0Ch .text:0040BE14 or eax, ecx .text:0040BE16 jnz short loc_40BE1C .text:0040BE18 mov bl, 1 .text:0040BE1A jmp short loc_40BE1E .text:0040BE1C ; --------------------------------------------------------------------------- .text:0040BE1C .text:0040BE1C loc_40BE1C: ; CODE XREF: start_search+176 j .text:0040BE1C xor bl, bl .text:0040BE1E .text:0040BE1E loc_40BE1E: ; CODE XREF: start_search+17A j .text:0040BE1E lea eax, [ebp+undo] .text:0040BE21 push eax .text:0040BE22 push esi .text:0040BE23 call unmake_move .text:0040BE28 add esp, 8 .text:0040BE2B test bl, bl .text:0040BE2D jz short loc_40BE39 .text:0040BE2F mov list_root[edi*4], esi .text:0040BE36 add edi, 1 ; num_moves++ .text:0040BE39 .text:0040BE39 loc_40BE39: ; CODE XREF: start_search+18D j .text:0040BE39 mov eax, [ebp+num_moves] .text:0040BE3C mov esi, [eax+4] .text:0040BE3F add eax, 4 ; i++ .text:0040BE42 test esi, esi ; list_root[i].move!=0 .text:0040BE44 mov [ebp+num_moves], eax .text:0040BE47 jnz short loc_40BDD8 .text:0040BE49 mov [ebp+num_moves], edi .text:0040BE4C .text:0040BE4C loc_40BE4C: ; CODE XREF: start_search+12F j .text:0040BE4C xor esi, esi .text:0040BE4E cmp edi, 1 ; if (num_moves <= 1) .text:0040BE51 mov list_root[edi*4], esi ; list_root[num_moves].move = 0 .text:0040BE58 ja short loc_40BE64 .text:0040BE5A mov G_max_depth, 4 ; max_depth = 4; .text:0040BE64 .text:0040BE64 loc_40BE64: ; CODE XREF: start_search+1B8 j .text:0040BE64 push esi .text:0040BE65 push offset dword_667A28 .text:0040BE6A call setjmp .text:0040BE6F add esp, 8 .text:0040BE72 test eax, eax ; if (setjmp(jmp_buffer) != 0) return; .text:0040BE74 jz short loc_40BE87 .text:0040BE76 mov eax, offset byte_664E08 .text:0040BE7B call sub_4092E0 .text:0040BE80 pop edi .text:0040BE81 pop esi .text:0040BE82 pop ebx .text:0040BE83 mov esp, ebp .text:0040BE85 pop ebp .text:0040BE86 retn .text:0040BE87 ; --------------------------------------------------------------------------- .text:0040BE87 .text:0040BE87 loc_40BE87: ; CODE XREF: start_search+1D4 j .text:0040BE87 call ds:GetTickCount ; start_timer .text:0040BE8D mov start_time, eax .text:0040BE92 mov eax, dword_66C490 .text:0040BE97 add eax, 1 ; All of this stuff is trans_inc_date .text:0040BE9A and eax, 3 .text:0040BE9D mov ecx, eax .text:0040BE9F cmp ecx, esi .text:0040BEA1 mov dword_66C490, eax .text:0040BEA6 jge short loc_40BEAB .text:0040BEA8 add ecx, 4 .text:0040BEAB .text:0040BEAB loc_40BEAB: ; CODE XREF: start_search+206 j .text:0040BEAB add ecx, 1 .text:0040BEAE shl ecx, 8 .text:0040BEB1 mov dword_66C498, ecx .text:0040BEB7 lea ecx, [eax-1] .text:0040BEBA cmp ecx, esi .text:0040BEBC jge short loc_40BEC1 .text:0040BEBE add ecx, 4 .text:0040BEC1 .text:0040BEC1 loc_40BEC1: ; CODE XREF: start_search+21C j .text:0040BEC1 add ecx, 1 .text:0040BEC4 shl ecx, 8 .text:0040BEC7 mov dword_66C49C, ecx .text:0040BECD lea ecx, [eax-2] .text:0040BED0 cmp ecx, esi .text:0040BED2 jge short loc_40BED7 .text:0040BED4 add ecx, 4 .text:0040BED7 .text:0040BED7 loc_40BED7: ; CODE XREF: start_search+232 j .text:0040BED7 add ecx, 1 .text:0040BEDA shl ecx, 8 .text:0040BEDD add eax, -3 .text:0040BEE0 cmp eax, esi .text:0040BEE2 mov dword_66C4A0, ecx .text:0040BEE8 jge short loc_40BEED .text:0040BEEA add eax, 4 .text:0040BEED .text:0040BEED loc_40BEED: ; CODE XREF: start_search+248 j .text:0040BEED add eax, 1 .text:0040BEF0 shl eax, 8 .text:0040BEF3 mov dword_66C4A4, eax .text:0040BEF8 xor eax, eax .text:0040BEFA lea ebx, [ebx+0] .text:0040BF00 .text:0040BF00 loc_40BF00: ; CODE XREF: start_search+274 j .text:0040BF00 mov killer1[eax], esi ; (Pos_info + i)->killer1 = 0; .text:0040BF06 mov killer2[eax], esi ; (Pos_info + i)->killer2 = 0; .text:0040BF0C add eax, 32 .text:0040BF0F cmp eax, 8192 .text:0040BF14 jb short loc_40BF00 .text:0040BF16 mov eax, 256 .text:0040BF1B mov ecx, 768 .text:0040BF20 mov edi, offset History .text:0040BF25 rep stosd ; memset(History, 256, 12 * 64 * sizeof(int)); .text:0040BF27 mov eax, dword_667C40 .text:0040BF2C and eax, dword_66C488 .text:0040BF32 mov ecx, dword_667C44 .text:0040BF38 shl eax, 4 .text:0040BF3B add eax, dword_66C480 .text:0040BF41 cmp [eax], ecx .text:0040BF43 jnz short loc_40BF4F .text:0040BF45 movzx edx, word ptr [eax+4] .text:0040BF49 mov dword_66C478, edx .text:0040BF4F .text:0040BF4F loc_40BF4F: ; CODE XREF: start_search+2A3 j .text:0040BF4F cmp [eax+10h], ecx .text:0040BF52 jnz short loc_40BF5E .text:0040BF54 movzx edx, word ptr [eax+14h] .text:0040BF58 mov dword_66C478, edx .text:0040BF5E .text:0040BF5E loc_40BF5E: ; CODE XREF: start_search+2B2 j .text:0040BF5E cmp [eax+20h], ecx .text:0040BF61 jnz short loc_40BF6D .text:0040BF63 movzx edx, word ptr [eax+24h] .text:0040BF67 mov dword_66C478, edx .text:0040BF6D .text:0040BF6D loc_40BF6D: ; CODE XREF: start_search+2C1 j .text:0040BF6D cmp [eax+30h], ecx .text:0040BF70 jnz short loc_40BF7B .text:0040BF72 movzx eax, word ptr [eax+34h] .text:0040BF76 mov dword_66C478, eax .text:0040BF7B .text:0040BF7B loc_40BF7B: ; CODE XREF: start_search+2D0 j .text:0040BF7B mov eax, [ebp+num_moves] .text:0040BF7E cmp eax, 2 .text:0040BF81 jb loc_40C039 .text:0040BF87 add eax, -1 .text:0040BF8A mov [ebp+var_8], eax .text:0040BF8D js loc_40C039 .text:0040BF93 .text:0040BF93 loc_40BF93: ; CODE XREF: start_search+393 j .text:0040BF93 lea ebx, ds:0[eax*4] .text:0040BF9A mov edi, list_root[ebx] .text:0040BFA0 cmp edi, dword_66C478 .text:0040BFA6 lea edx, dword_66C074[ebx] .text:0040BFAC jnz short loc_40BFB5 .text:0040BFAE mov ecx, 7FFFFFFFh .text:0040BFB3 jmp short loc_40BFE1 .text:0040BFB5 ; --------------------------------------------------------------------------- .text:0040BFB5 .text:0040BFB5 loc_40BFB5: ; CODE XREF: start_search+30C j .text:0040BFB5 mov ecx, edi .text:0040BFB7 and ecx, 63 .text:0040BFBA mov ecx, dword_667A90[ecx*4] .text:0040BFC1 test ecx, ecx .text:0040BFC3 jz short loc_40BFDF .text:0040BFC5 mov esi, edi .text:0040BFC7 shr esi, 6 .text:0040BFCA and esi, 63 .text:0040BFCD shl ecx, 4 .text:0040BFD0 sub ecx, dword_667A90[esi*4] .text:0040BFD7 add ecx, 2147483391 .text:0040BFDD jmp short loc_40BFE1 .text:0040BFDF ; --------------------------------------------------------------------------- .text:0040BFDF .text:0040BFDF loc_40BFDF: ; CODE XREF: start_search+323 j .text:0040BFDF xor ecx, ecx .text:0040BFE1 .text:0040BFE1 loc_40BFE1: ; CODE XREF: start_search+313 j .text:0040BFE1 ; start_search+33D j .text:0040BFE1 cmp ecx, [edx] .text:0040BFE3 mov esi, eax .text:0040BFE5 jnb short loc_40C01F .text:0040BFE7 mov eax, ebx .text:0040BFE9 lea esp, [esp+0] .text:0040BFF0 .text:0040BFF0 loc_40BFF0: ; CODE XREF: start_search+37A j .text:0040BFF0 mov ebx, dword_66BC74[eax] .text:0040BFF6 mov edx, [edx] .text:0040BFF8 mov list_root[eax], ebx .text:0040BFFE mov dword_66C070[eax], edx .text:0040C004 add esi, 1 .text:0040C007 lea eax, ds:0[esi*4] .text:0040C00E cmp ecx, dword_66C074[eax] .text:0040C014 lea edx, dword_66C074[eax] .text:0040C01A jb short loc_40BFF0 .text:0040C01C mov eax, [ebp+var_8] .text:0040C01F .text:0040C01F loc_40C01F: ; CODE XREF: start_search+345 j .text:0040C01F sub eax, 1 .text:0040C022 mov list_root[esi*4], edi .text:0040C029 mov dword_66C070[esi*4], ecx .text:0040C030 mov [ebp+var_8], eax .text:0040C033 jns loc_40BF93 .text:0040C039 .text:0040C039 loc_40C039: ; CODE XREF: start_search+2E1 j .text:0040C039 ; start_search+2ED j
--Setjmp is called after the move generation, as in Fruit
--Something very strange: whereas in Strelka a failed setjmp call simply results in a return, in Rybka 1.0 it involves a function call, like in Fruit (there it is search_update_current()). The weird part is, it appears to be calling parse_position(). That is what the "call sub_4092E0" is for. Why would Rybka 1.0 be parsing the position after a setjmp call?
--The timer is started after the setjmp call
--depth is limited to 4 when there's only 1 legal move (commented out in Strelka, as noted in the other thread).
--I noted earlier that the history initialization was different. Strelka and Rybka 1.0 are the same here, using memset(History, 256, 12 * 64 * sizeof(int)); But what I didn't notice is, 256 is out of range for an unsigned char. So that code is the same as Fruit's memset(History, 0, 12 * 64 * sizeof(int)); Why would Rybka 1.0 have 256? Is it the optimizer gone wild, or a deliberate attempt at obfuscation?
-
bob
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
-
chrisw
Re: Wanted: some opposition to the provided evidence
Well, I can see a bunch of assembler mnemonics, then what? It's Rybka? Where's the Fruit? Then I can see five points in english text of which the last point has had to be withdrawn on challenge.bob wrote:Did you see Z's post doing just that???chrisw wrote:From Rybka forum, Vas states:bob wrote:The only possible "innocent explanation" would be "OK, I started by modifying fruit and I have modified the important parts, but didn't realize that the other parts would represent a problem". If you take any two chess programs on the planet, and compare the source, you are not going to find even a hundred lines of code that match exactly, much less a thousand or more. Unless they have a common ancestry, that is.chrisw wrote:What if there's a completely innocent explanation?kranium wrote:No, not everything is speculative.Enir wrote:Maybe. Reverse engineering Rybka 3 will prove whether or not it contains GPL code. Everything else is just speculative and damaging.bnemias wrote:If you need examples, Verizon sued by BusyBox developers. Redhat v SCO, Sun v Microsoft. These occur all the time because major players push software that isn't "squeaky clean." So it's hard to use Chessbase's intelligence as proof of anything concerning Rybka 3.tiger wrote:I think I have first hand experience on this and I know that you are wrong.chrisw wrote:[snip]
No commercial company with half a brain will publish software claiming total rights over it if not true. That's the overpowering evidence one that R3 is clean. Chessbase won't licence it unless absolutely certain it's clean. That's overpowering evidence two. Say anything else and you play with fire. imo.
Enrique
There's a mountain of evidence that strleka 2.0 has hundreds and hundreds of lines of code that are identical or equivilent to code found in fruit 2.1. (this comparison can be made even by a casual user...the two sets of source code are freely available). a detailed comparison was presented here:
http://64.68.157.89/forum/viewtopic.php ... ht=strelka
There is also evidence that strelka 2.0 (both source code and binary) is identical to rybka 1.0 beta.
In addition Vas has confirmed that Strelka is indeed his.
? I think the conclusion is not difficult to ascertain...
rybka 1.0 beta is a derivative of fruit 2.1
i volunteer to stir the shit....
Maybe somebody can summarize the points.
Generally, code theft is easy to show - just show the two sections of identical code, side-by-side. There isn't much to debate in such cases. Rybka is of course original (with some accepted exceptions like bit scans & bit counters, etc). Strelka contains Rybka code. Whether Strelka also contains Fruit code, I don't know and don't really care.
If you want to show a problem with Rybka, put it, Rybka side by side with Fruit.
So exactly what is Zach's listing saying when translated from Egyptian hieroglyphics?
The judge will ask you to go away and come back with both Fruit and Rybka side by side so he can see the similarities for himself.
He'll also ask the other side to comment. We've not seen this yet either.