Decoding assembly instructions in a Game Boy disassembler












8












$begingroup$


I am coding a game boy disassembler in C++. The program (simplified here) converts a vector of byte into an assembly instruction and print it.



To do so, it iterates through the vector of bytes (char) with an iterator. I noticed that with the it!=char_vect.end() condition, there is the risk that the iterator goes out from the vector. So I have to check that I do not get out of the vector at each iteration and if it occurs, I throw an error.



#include <iostream>
#include <vector>

int disassemble(std::vector<char>::iterator& it){

int opcode = *it;
int opbytes = 1; //number of bytes used by the operator

switch(opcode){
case 0x76 : std::cout<<"HALT "<<std::endl; opbytes = 1; break;
case 0x10 : std::cout<<"STOP $"<<(int)*(it+1)<<std::endl; opbytes =2; break;
//... About 250 different cases
default : std::cout<<"Instruction not implemented yet"<<std::endl;
}

return opbytes;
}

int main(){

std::vector<char> char_vect = {0x76, 0x10, 0x20}; // Fill vector of char
std::vector<char>::iterator it = char_vect.begin();

// First I only did that
//while (it!=char_vect.end()){
// it+=disassemble(it);
//}

// But I have to check that I do not go out of the char_vect

while (it!=char_vect.end()){
try{
it+=disassemble(it);
//check that it do not overpass end
if (size_t(it - char_vect.begin())>char_vect.size()){
throw "Try to disassemble instruction outside of memory bounds";
}
}
catch(const char * c){
std::cerr << "Fatal error: " << c << std::endl;
return 1;
}
}
return 1;
}


This code works fine. Now I wonder, I have a possible exception case which is the iterator going further than vect_char.end(). Is a try / catch an appropriate way to handle this case knowing that:





  1. The reason for the iterator to go out of bounds could be either:




    • A non valid input byte sequence, for example {0x76, 0x20, 0x10} since the instruction 0x10 expects an argument. This should not append.

    • A mistake in the disassemble function. For exemple with the valid input {0x76, 0x10, 0x20}, if I code by mistake that 0x10 uses 3 opbytes instead of 2, the iterator will go out of bounds. This should not append either.



  2. If I iterated through the vect_char with an index, I would not have this issue and the code would be more compact



I never really used try/catch before and I do not know if it is intented for such situations, so I would like to know, how would you have handled this unexpected case?










share|improve this question









New contributor




Mai Kar is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.







$endgroup$








  • 3




    $begingroup$
    This question might be more on-topic over at SoftwareEngineering.StackExchange.com . However, before posting please follow their tour and read "How do I ask a good question?", "What topics can I ask about here?" and "What types of questions should I avoid asking?".
    $endgroup$
    – BCdotWEB
    15 hours ago










  • $begingroup$
    Does or doesn't the current code work as intended? I assume you've tested it?
    $endgroup$
    – Mast
    14 hours ago






  • 1




    $begingroup$
    @Mast Yes it does work as intended, I just tested it.
    $endgroup$
    – Mai Kar
    14 hours ago






  • 2




    $begingroup$
    Your error is simply passing insufficient Information to disassemble().
    $endgroup$
    – Deduplicator
    13 hours ago






  • 3




    $begingroup$
    I reformulated my demand because I think it was not clear enough. This question is mainly about validating my design choices, i.e. the use of iterators and try/catch.
    $endgroup$
    – Mai Kar
    12 hours ago
















8












$begingroup$


I am coding a game boy disassembler in C++. The program (simplified here) converts a vector of byte into an assembly instruction and print it.



To do so, it iterates through the vector of bytes (char) with an iterator. I noticed that with the it!=char_vect.end() condition, there is the risk that the iterator goes out from the vector. So I have to check that I do not get out of the vector at each iteration and if it occurs, I throw an error.



#include <iostream>
#include <vector>

int disassemble(std::vector<char>::iterator& it){

int opcode = *it;
int opbytes = 1; //number of bytes used by the operator

switch(opcode){
case 0x76 : std::cout<<"HALT "<<std::endl; opbytes = 1; break;
case 0x10 : std::cout<<"STOP $"<<(int)*(it+1)<<std::endl; opbytes =2; break;
//... About 250 different cases
default : std::cout<<"Instruction not implemented yet"<<std::endl;
}

return opbytes;
}

int main(){

std::vector<char> char_vect = {0x76, 0x10, 0x20}; // Fill vector of char
std::vector<char>::iterator it = char_vect.begin();

// First I only did that
//while (it!=char_vect.end()){
// it+=disassemble(it);
//}

// But I have to check that I do not go out of the char_vect

while (it!=char_vect.end()){
try{
it+=disassemble(it);
//check that it do not overpass end
if (size_t(it - char_vect.begin())>char_vect.size()){
throw "Try to disassemble instruction outside of memory bounds";
}
}
catch(const char * c){
std::cerr << "Fatal error: " << c << std::endl;
return 1;
}
}
return 1;
}


This code works fine. Now I wonder, I have a possible exception case which is the iterator going further than vect_char.end(). Is a try / catch an appropriate way to handle this case knowing that:





  1. The reason for the iterator to go out of bounds could be either:




    • A non valid input byte sequence, for example {0x76, 0x20, 0x10} since the instruction 0x10 expects an argument. This should not append.

    • A mistake in the disassemble function. For exemple with the valid input {0x76, 0x10, 0x20}, if I code by mistake that 0x10 uses 3 opbytes instead of 2, the iterator will go out of bounds. This should not append either.



  2. If I iterated through the vect_char with an index, I would not have this issue and the code would be more compact



I never really used try/catch before and I do not know if it is intented for such situations, so I would like to know, how would you have handled this unexpected case?










share|improve this question









New contributor




Mai Kar is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.







$endgroup$








  • 3




    $begingroup$
    This question might be more on-topic over at SoftwareEngineering.StackExchange.com . However, before posting please follow their tour and read "How do I ask a good question?", "What topics can I ask about here?" and "What types of questions should I avoid asking?".
    $endgroup$
    – BCdotWEB
    15 hours ago










  • $begingroup$
    Does or doesn't the current code work as intended? I assume you've tested it?
    $endgroup$
    – Mast
    14 hours ago






  • 1




    $begingroup$
    @Mast Yes it does work as intended, I just tested it.
    $endgroup$
    – Mai Kar
    14 hours ago






  • 2




    $begingroup$
    Your error is simply passing insufficient Information to disassemble().
    $endgroup$
    – Deduplicator
    13 hours ago






  • 3




    $begingroup$
    I reformulated my demand because I think it was not clear enough. This question is mainly about validating my design choices, i.e. the use of iterators and try/catch.
    $endgroup$
    – Mai Kar
    12 hours ago














8












8








8





$begingroup$


I am coding a game boy disassembler in C++. The program (simplified here) converts a vector of byte into an assembly instruction and print it.



To do so, it iterates through the vector of bytes (char) with an iterator. I noticed that with the it!=char_vect.end() condition, there is the risk that the iterator goes out from the vector. So I have to check that I do not get out of the vector at each iteration and if it occurs, I throw an error.



#include <iostream>
#include <vector>

int disassemble(std::vector<char>::iterator& it){

int opcode = *it;
int opbytes = 1; //number of bytes used by the operator

switch(opcode){
case 0x76 : std::cout<<"HALT "<<std::endl; opbytes = 1; break;
case 0x10 : std::cout<<"STOP $"<<(int)*(it+1)<<std::endl; opbytes =2; break;
//... About 250 different cases
default : std::cout<<"Instruction not implemented yet"<<std::endl;
}

return opbytes;
}

int main(){

std::vector<char> char_vect = {0x76, 0x10, 0x20}; // Fill vector of char
std::vector<char>::iterator it = char_vect.begin();

// First I only did that
//while (it!=char_vect.end()){
// it+=disassemble(it);
//}

// But I have to check that I do not go out of the char_vect

while (it!=char_vect.end()){
try{
it+=disassemble(it);
//check that it do not overpass end
if (size_t(it - char_vect.begin())>char_vect.size()){
throw "Try to disassemble instruction outside of memory bounds";
}
}
catch(const char * c){
std::cerr << "Fatal error: " << c << std::endl;
return 1;
}
}
return 1;
}


This code works fine. Now I wonder, I have a possible exception case which is the iterator going further than vect_char.end(). Is a try / catch an appropriate way to handle this case knowing that:





  1. The reason for the iterator to go out of bounds could be either:




    • A non valid input byte sequence, for example {0x76, 0x20, 0x10} since the instruction 0x10 expects an argument. This should not append.

    • A mistake in the disassemble function. For exemple with the valid input {0x76, 0x10, 0x20}, if I code by mistake that 0x10 uses 3 opbytes instead of 2, the iterator will go out of bounds. This should not append either.



  2. If I iterated through the vect_char with an index, I would not have this issue and the code would be more compact



I never really used try/catch before and I do not know if it is intented for such situations, so I would like to know, how would you have handled this unexpected case?










share|improve this question









New contributor




Mai Kar is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.







$endgroup$




I am coding a game boy disassembler in C++. The program (simplified here) converts a vector of byte into an assembly instruction and print it.



To do so, it iterates through the vector of bytes (char) with an iterator. I noticed that with the it!=char_vect.end() condition, there is the risk that the iterator goes out from the vector. So I have to check that I do not get out of the vector at each iteration and if it occurs, I throw an error.



#include <iostream>
#include <vector>

int disassemble(std::vector<char>::iterator& it){

int opcode = *it;
int opbytes = 1; //number of bytes used by the operator

switch(opcode){
case 0x76 : std::cout<<"HALT "<<std::endl; opbytes = 1; break;
case 0x10 : std::cout<<"STOP $"<<(int)*(it+1)<<std::endl; opbytes =2; break;
//... About 250 different cases
default : std::cout<<"Instruction not implemented yet"<<std::endl;
}

return opbytes;
}

int main(){

std::vector<char> char_vect = {0x76, 0x10, 0x20}; // Fill vector of char
std::vector<char>::iterator it = char_vect.begin();

// First I only did that
//while (it!=char_vect.end()){
// it+=disassemble(it);
//}

// But I have to check that I do not go out of the char_vect

while (it!=char_vect.end()){
try{
it+=disassemble(it);
//check that it do not overpass end
if (size_t(it - char_vect.begin())>char_vect.size()){
throw "Try to disassemble instruction outside of memory bounds";
}
}
catch(const char * c){
std::cerr << "Fatal error: " << c << std::endl;
return 1;
}
}
return 1;
}


This code works fine. Now I wonder, I have a possible exception case which is the iterator going further than vect_char.end(). Is a try / catch an appropriate way to handle this case knowing that:





  1. The reason for the iterator to go out of bounds could be either:




    • A non valid input byte sequence, for example {0x76, 0x20, 0x10} since the instruction 0x10 expects an argument. This should not append.

    • A mistake in the disassemble function. For exemple with the valid input {0x76, 0x10, 0x20}, if I code by mistake that 0x10 uses 3 opbytes instead of 2, the iterator will go out of bounds. This should not append either.



  2. If I iterated through the vect_char with an index, I would not have this issue and the code would be more compact



I never really used try/catch before and I do not know if it is intented for such situations, so I would like to know, how would you have handled this unexpected case?







c++ error-handling iterator assembly serialization






share|improve this question









New contributor




Mai Kar is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.











share|improve this question









New contributor




Mai Kar is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.









share|improve this question




share|improve this question








edited 12 hours ago







Mai Kar













New contributor




Mai Kar is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.









asked 15 hours ago









Mai KarMai Kar

465




465




New contributor




Mai Kar is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.





New contributor





Mai Kar is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.






Mai Kar is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.








  • 3




    $begingroup$
    This question might be more on-topic over at SoftwareEngineering.StackExchange.com . However, before posting please follow their tour and read "How do I ask a good question?", "What topics can I ask about here?" and "What types of questions should I avoid asking?".
    $endgroup$
    – BCdotWEB
    15 hours ago










  • $begingroup$
    Does or doesn't the current code work as intended? I assume you've tested it?
    $endgroup$
    – Mast
    14 hours ago






  • 1




    $begingroup$
    @Mast Yes it does work as intended, I just tested it.
    $endgroup$
    – Mai Kar
    14 hours ago






  • 2




    $begingroup$
    Your error is simply passing insufficient Information to disassemble().
    $endgroup$
    – Deduplicator
    13 hours ago






  • 3




    $begingroup$
    I reformulated my demand because I think it was not clear enough. This question is mainly about validating my design choices, i.e. the use of iterators and try/catch.
    $endgroup$
    – Mai Kar
    12 hours ago














  • 3




    $begingroup$
    This question might be more on-topic over at SoftwareEngineering.StackExchange.com . However, before posting please follow their tour and read "How do I ask a good question?", "What topics can I ask about here?" and "What types of questions should I avoid asking?".
    $endgroup$
    – BCdotWEB
    15 hours ago










  • $begingroup$
    Does or doesn't the current code work as intended? I assume you've tested it?
    $endgroup$
    – Mast
    14 hours ago






  • 1




    $begingroup$
    @Mast Yes it does work as intended, I just tested it.
    $endgroup$
    – Mai Kar
    14 hours ago






  • 2




    $begingroup$
    Your error is simply passing insufficient Information to disassemble().
    $endgroup$
    – Deduplicator
    13 hours ago






  • 3




    $begingroup$
    I reformulated my demand because I think it was not clear enough. This question is mainly about validating my design choices, i.e. the use of iterators and try/catch.
    $endgroup$
    – Mai Kar
    12 hours ago








3




3




$begingroup$
This question might be more on-topic over at SoftwareEngineering.StackExchange.com . However, before posting please follow their tour and read "How do I ask a good question?", "What topics can I ask about here?" and "What types of questions should I avoid asking?".
$endgroup$
– BCdotWEB
15 hours ago




$begingroup$
This question might be more on-topic over at SoftwareEngineering.StackExchange.com . However, before posting please follow their tour and read "How do I ask a good question?", "What topics can I ask about here?" and "What types of questions should I avoid asking?".
$endgroup$
– BCdotWEB
15 hours ago












$begingroup$
Does or doesn't the current code work as intended? I assume you've tested it?
$endgroup$
– Mast
14 hours ago




$begingroup$
Does or doesn't the current code work as intended? I assume you've tested it?
$endgroup$
– Mast
14 hours ago




1




1




$begingroup$
@Mast Yes it does work as intended, I just tested it.
$endgroup$
– Mai Kar
14 hours ago




$begingroup$
@Mast Yes it does work as intended, I just tested it.
$endgroup$
– Mai Kar
14 hours ago




2




2




$begingroup$
Your error is simply passing insufficient Information to disassemble().
$endgroup$
– Deduplicator
13 hours ago




$begingroup$
Your error is simply passing insufficient Information to disassemble().
$endgroup$
– Deduplicator
13 hours ago




3




3




$begingroup$
I reformulated my demand because I think it was not clear enough. This question is mainly about validating my design choices, i.e. the use of iterators and try/catch.
$endgroup$
– Mai Kar
12 hours ago




$begingroup$
I reformulated my demand because I think it was not clear enough. This question is mainly about validating my design choices, i.e. the use of iterators and try/catch.
$endgroup$
– Mai Kar
12 hours ago










3 Answers
3






active

oldest

votes


















8












$begingroup$

I have some ideas about how you might be able to improve your program.



Avoid problems



Rather than trying to deal with the problem for every instruction, one approach is avoiding it entirely. One way to do that is to simply append a number of bytes to the end of the vector. If the maximum bytes for an instruction is $n$, then append $n-1$ bytes to the end and stop when you've advanced into the padded area.



Check before advancing



One could also pass the number of remaining bytes to the disassemble function. However, the mechanism I'd suggest would be to pass a range, e.g.:



int diassembleOne(std::vector<char>::iterator& it, std::vector<char>::iterator& end) {
// figure out number of bytes for this opcode
if (std::distance(it, end) > opbytes) {
it = end;
// throw or return 0
}
// disassemble the instruction thoroughly
std::advance(it, opbytes);
return opbytes;
}


Use const iterators



If all the code is doing is disassembling, then it shouldn't alter the underlying vector. For that reason, I'd recommend passing a std::vector<char>::const_iterator &.



Use classes



I'd suggest using an Opcode class like this:



class Opcode {
char code;
short int bytes;
std::string_view name;
bool operator==(char opcode) const { return code == opcode; }
int decode(std::vector<char>::const_iterator& it, std::ostream& out=std::cout) const {
out << name;
++it;
for (int i{bytes-1}; i; --i) {
out << static_cast<unsigned>(*it++);
}
out << 'n';
return bytes;
}
};

constexpr std::array<Opcode,2> instructions {
{ 0x10, 2, "STOP $" },
{ 0x76, 2, "HALT " },
};


Pass a pair of iterators to the dissemble function



As mentioned before, you can pass a pair of iterators to the disassemble function. Using that plus the class above:



int disassembleOne(std::vector<char>::const_iterator& it, std::vector<char>::const_iterator& end){
auto op{std::find(instructions.begin(), instructions.end(), *it)};
if (op == instructions.end()) {
std::cout << "Instruction not foundn";
it = end;
return 0; // instruction not found or off the end of the passed array
}
if (std::distance(it, end) < op->bytes) {
std::cout << "Not enough bytes left to decode " << op->name << 'n';
it = end;
return 0; // instruction not found or off the end of the passed array
}
return op->decode(it);
}


Now main becomes very simple:



int main(){
const std::vector<char> char_vect = {0x76, 0x10, 0x20, 0x10}; // Fill vector of char
auto end{char_vect.cend()};

for (auto it{char_vect.cbegin()}; it != end; disassembleOne(it, end)) {
}
}


Another way to do this would be to put more of the processing in the Opcode itself -- it would make sense that each opcode would know how to decode itself.



Be clear about caller expectations



This code, as with the original code, expects that the it being passed is a valid iterator that is not the end. It is good to document that in comments in the code.






share|improve this answer











$endgroup$













  • $begingroup$
    The loop in main doesn't use the result of the call to disassembleOne.
    $endgroup$
    – Roland Illig
    9 hours ago










  • $begingroup$
    @RolandIllig It doesn't need to use the return value because the iterator, which is passed by reference, is updated within the function.
    $endgroup$
    – Edward
    8 hours ago



















4












$begingroup$

I think the big switch is a problem. Consider a more data-driven approach where each opcode is described by a struct:



struct OpCode
{
unsigned char command_byte;
unsigned char mask; // if only a subset of bits determine command
unsigned char length;
// other members as needed - e.g. a pointer to the "print" function
};


Now, the code that reads instructions can determine whether the command is unterminated, without needing to repeat the logic for every multi-byte opcode.



I've included the mask so that we don't encode every single instruction (e.g. ld R1, R2 encodes the source and destination registers in the bits of a single-byte command; it would be tedious and error-prone to write them all separately here).



The simple length value I've shown isn't quite enough, given that the Game Boy's LR35902 processor supports the 0xCB prefix for extended Z80 instructions - we might want to handle that outside of the normal flow, and use it to switch between different instruction tables.

We get away with a simple length value here, because the only prefix instruction supported by the Game Boy's LR35902 processor is 0xCB, which is always followed by a single-byte instruction. If we were decoding Z80 instructions (with ED prefix), then we'd need something a little more sophisticated.






share|improve this answer









$endgroup$





















    2












    $begingroup$

    Your code's division into functions isn't very natural. The objective is to have each function performing one specific task, which should also be as orthogonal as possible to the tasks performed by the other functions. For instance, your disassemble function performs three different functions: it reads from the instruction stream, it interprets assembly code and returns the number of bytes that should be skipped to get to the next instruction. That's a not-so-coherent mix of responsibilities.



    There's also a bug in your code, because it+=disassemble(it); could point beyond char_vect.end() which is in itself undefined behavior, even if you don't dereference the iterator.



    I would build my disassembler around an iterator, or better even a range (if you don't mind external libraries or anticipating the next standard):



    #include <algorithm>
    #include <vector>
    #include <iostream>
    #include <range/v3/view.hpp>

    using iterator = std::vector<char>::const_iterator;
    struct truncated_instruction {};

    // one function to interpret the instruction
    void interpret(char instruction) {
    if (instruction == 'x' throw bad_instruction();
    std::cout << (int) instruction << ' ';
    }

    // one function to get the number of bytes to skip
    int nb_bytes(char c) { return 1; }

    class disassembled
    : public ranges::view_facade<disassembled>
    {
    friend ranges::range_access;
    iterator first, last;
    char const & read() const { return *first; }
    bool equal(ranges::default_sentinel) const { return first == last; }
    void next() {
    // one function to get to the next instruction
    auto bytes_to_skip = nb_bytes(*first);
    if (std::distance(first, last) < bytes_to_skip) throw truncated_instruction();
    // check if there's enough space left to advance before advancing
    std::advance(first, bytes_to_skip);
    }
    public:
    disassembled() = default;
    explicit disassembled(const std::vector<char>& v) : first(v.begin()), last(v.end()) {}
    };

    int main() {
    std::vector<char> char_vect = {0x76, 0x10, 0x20, 0x30};
    try {
    for (auto instruction : disassembled(char_vect)) interpret(instruction);
    }
    catch // ...
    }





    share|improve this answer









    $endgroup$













      Your Answer





      StackExchange.ifUsing("editor", function () {
      return StackExchange.using("mathjaxEditing", function () {
      StackExchange.MarkdownEditor.creationCallbacks.add(function (editor, postfix) {
      StackExchange.mathjaxEditing.prepareWmdForMathJax(editor, postfix, [["\$", "\$"]]);
      });
      });
      }, "mathjax-editing");

      StackExchange.ifUsing("editor", function () {
      StackExchange.using("externalEditor", function () {
      StackExchange.using("snippets", function () {
      StackExchange.snippets.init();
      });
      });
      }, "code-snippets");

      StackExchange.ready(function() {
      var channelOptions = {
      tags: "".split(" "),
      id: "196"
      };
      initTagRenderer("".split(" "), "".split(" "), channelOptions);

      StackExchange.using("externalEditor", function() {
      // Have to fire editor after snippets, if snippets enabled
      if (StackExchange.settings.snippets.snippetsEnabled) {
      StackExchange.using("snippets", function() {
      createEditor();
      });
      }
      else {
      createEditor();
      }
      });

      function createEditor() {
      StackExchange.prepareEditor({
      heartbeatType: 'answer',
      autoActivateHeartbeat: false,
      convertImagesToLinks: false,
      noModals: true,
      showLowRepImageUploadWarning: true,
      reputationToPostImages: null,
      bindNavPrevention: true,
      postfix: "",
      imageUploader: {
      brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
      contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
      allowUrls: true
      },
      onDemand: true,
      discardSelector: ".discard-answer"
      ,immediatelyShowMarkdownHelp:true
      });


      }
      });






      Mai Kar is a new contributor. Be nice, and check out our Code of Conduct.










      draft saved

      draft discarded


















      StackExchange.ready(
      function () {
      StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fcodereview.stackexchange.com%2fquestions%2f215349%2fdecoding-assembly-instructions-in-a-game-boy-disassembler%23new-answer', 'question_page');
      }
      );

      Post as a guest















      Required, but never shown

























      3 Answers
      3






      active

      oldest

      votes








      3 Answers
      3






      active

      oldest

      votes









      active

      oldest

      votes






      active

      oldest

      votes









      8












      $begingroup$

      I have some ideas about how you might be able to improve your program.



      Avoid problems



      Rather than trying to deal with the problem for every instruction, one approach is avoiding it entirely. One way to do that is to simply append a number of bytes to the end of the vector. If the maximum bytes for an instruction is $n$, then append $n-1$ bytes to the end and stop when you've advanced into the padded area.



      Check before advancing



      One could also pass the number of remaining bytes to the disassemble function. However, the mechanism I'd suggest would be to pass a range, e.g.:



      int diassembleOne(std::vector<char>::iterator& it, std::vector<char>::iterator& end) {
      // figure out number of bytes for this opcode
      if (std::distance(it, end) > opbytes) {
      it = end;
      // throw or return 0
      }
      // disassemble the instruction thoroughly
      std::advance(it, opbytes);
      return opbytes;
      }


      Use const iterators



      If all the code is doing is disassembling, then it shouldn't alter the underlying vector. For that reason, I'd recommend passing a std::vector<char>::const_iterator &.



      Use classes



      I'd suggest using an Opcode class like this:



      class Opcode {
      char code;
      short int bytes;
      std::string_view name;
      bool operator==(char opcode) const { return code == opcode; }
      int decode(std::vector<char>::const_iterator& it, std::ostream& out=std::cout) const {
      out << name;
      ++it;
      for (int i{bytes-1}; i; --i) {
      out << static_cast<unsigned>(*it++);
      }
      out << 'n';
      return bytes;
      }
      };

      constexpr std::array<Opcode,2> instructions {
      { 0x10, 2, "STOP $" },
      { 0x76, 2, "HALT " },
      };


      Pass a pair of iterators to the dissemble function



      As mentioned before, you can pass a pair of iterators to the disassemble function. Using that plus the class above:



      int disassembleOne(std::vector<char>::const_iterator& it, std::vector<char>::const_iterator& end){
      auto op{std::find(instructions.begin(), instructions.end(), *it)};
      if (op == instructions.end()) {
      std::cout << "Instruction not foundn";
      it = end;
      return 0; // instruction not found or off the end of the passed array
      }
      if (std::distance(it, end) < op->bytes) {
      std::cout << "Not enough bytes left to decode " << op->name << 'n';
      it = end;
      return 0; // instruction not found or off the end of the passed array
      }
      return op->decode(it);
      }


      Now main becomes very simple:



      int main(){
      const std::vector<char> char_vect = {0x76, 0x10, 0x20, 0x10}; // Fill vector of char
      auto end{char_vect.cend()};

      for (auto it{char_vect.cbegin()}; it != end; disassembleOne(it, end)) {
      }
      }


      Another way to do this would be to put more of the processing in the Opcode itself -- it would make sense that each opcode would know how to decode itself.



      Be clear about caller expectations



      This code, as with the original code, expects that the it being passed is a valid iterator that is not the end. It is good to document that in comments in the code.






      share|improve this answer











      $endgroup$













      • $begingroup$
        The loop in main doesn't use the result of the call to disassembleOne.
        $endgroup$
        – Roland Illig
        9 hours ago










      • $begingroup$
        @RolandIllig It doesn't need to use the return value because the iterator, which is passed by reference, is updated within the function.
        $endgroup$
        – Edward
        8 hours ago
















      8












      $begingroup$

      I have some ideas about how you might be able to improve your program.



      Avoid problems



      Rather than trying to deal with the problem for every instruction, one approach is avoiding it entirely. One way to do that is to simply append a number of bytes to the end of the vector. If the maximum bytes for an instruction is $n$, then append $n-1$ bytes to the end and stop when you've advanced into the padded area.



      Check before advancing



      One could also pass the number of remaining bytes to the disassemble function. However, the mechanism I'd suggest would be to pass a range, e.g.:



      int diassembleOne(std::vector<char>::iterator& it, std::vector<char>::iterator& end) {
      // figure out number of bytes for this opcode
      if (std::distance(it, end) > opbytes) {
      it = end;
      // throw or return 0
      }
      // disassemble the instruction thoroughly
      std::advance(it, opbytes);
      return opbytes;
      }


      Use const iterators



      If all the code is doing is disassembling, then it shouldn't alter the underlying vector. For that reason, I'd recommend passing a std::vector<char>::const_iterator &.



      Use classes



      I'd suggest using an Opcode class like this:



      class Opcode {
      char code;
      short int bytes;
      std::string_view name;
      bool operator==(char opcode) const { return code == opcode; }
      int decode(std::vector<char>::const_iterator& it, std::ostream& out=std::cout) const {
      out << name;
      ++it;
      for (int i{bytes-1}; i; --i) {
      out << static_cast<unsigned>(*it++);
      }
      out << 'n';
      return bytes;
      }
      };

      constexpr std::array<Opcode,2> instructions {
      { 0x10, 2, "STOP $" },
      { 0x76, 2, "HALT " },
      };


      Pass a pair of iterators to the dissemble function



      As mentioned before, you can pass a pair of iterators to the disassemble function. Using that plus the class above:



      int disassembleOne(std::vector<char>::const_iterator& it, std::vector<char>::const_iterator& end){
      auto op{std::find(instructions.begin(), instructions.end(), *it)};
      if (op == instructions.end()) {
      std::cout << "Instruction not foundn";
      it = end;
      return 0; // instruction not found or off the end of the passed array
      }
      if (std::distance(it, end) < op->bytes) {
      std::cout << "Not enough bytes left to decode " << op->name << 'n';
      it = end;
      return 0; // instruction not found or off the end of the passed array
      }
      return op->decode(it);
      }


      Now main becomes very simple:



      int main(){
      const std::vector<char> char_vect = {0x76, 0x10, 0x20, 0x10}; // Fill vector of char
      auto end{char_vect.cend()};

      for (auto it{char_vect.cbegin()}; it != end; disassembleOne(it, end)) {
      }
      }


      Another way to do this would be to put more of the processing in the Opcode itself -- it would make sense that each opcode would know how to decode itself.



      Be clear about caller expectations



      This code, as with the original code, expects that the it being passed is a valid iterator that is not the end. It is good to document that in comments in the code.






      share|improve this answer











      $endgroup$













      • $begingroup$
        The loop in main doesn't use the result of the call to disassembleOne.
        $endgroup$
        – Roland Illig
        9 hours ago










      • $begingroup$
        @RolandIllig It doesn't need to use the return value because the iterator, which is passed by reference, is updated within the function.
        $endgroup$
        – Edward
        8 hours ago














      8












      8








      8





      $begingroup$

      I have some ideas about how you might be able to improve your program.



      Avoid problems



      Rather than trying to deal with the problem for every instruction, one approach is avoiding it entirely. One way to do that is to simply append a number of bytes to the end of the vector. If the maximum bytes for an instruction is $n$, then append $n-1$ bytes to the end and stop when you've advanced into the padded area.



      Check before advancing



      One could also pass the number of remaining bytes to the disassemble function. However, the mechanism I'd suggest would be to pass a range, e.g.:



      int diassembleOne(std::vector<char>::iterator& it, std::vector<char>::iterator& end) {
      // figure out number of bytes for this opcode
      if (std::distance(it, end) > opbytes) {
      it = end;
      // throw or return 0
      }
      // disassemble the instruction thoroughly
      std::advance(it, opbytes);
      return opbytes;
      }


      Use const iterators



      If all the code is doing is disassembling, then it shouldn't alter the underlying vector. For that reason, I'd recommend passing a std::vector<char>::const_iterator &.



      Use classes



      I'd suggest using an Opcode class like this:



      class Opcode {
      char code;
      short int bytes;
      std::string_view name;
      bool operator==(char opcode) const { return code == opcode; }
      int decode(std::vector<char>::const_iterator& it, std::ostream& out=std::cout) const {
      out << name;
      ++it;
      for (int i{bytes-1}; i; --i) {
      out << static_cast<unsigned>(*it++);
      }
      out << 'n';
      return bytes;
      }
      };

      constexpr std::array<Opcode,2> instructions {
      { 0x10, 2, "STOP $" },
      { 0x76, 2, "HALT " },
      };


      Pass a pair of iterators to the dissemble function



      As mentioned before, you can pass a pair of iterators to the disassemble function. Using that plus the class above:



      int disassembleOne(std::vector<char>::const_iterator& it, std::vector<char>::const_iterator& end){
      auto op{std::find(instructions.begin(), instructions.end(), *it)};
      if (op == instructions.end()) {
      std::cout << "Instruction not foundn";
      it = end;
      return 0; // instruction not found or off the end of the passed array
      }
      if (std::distance(it, end) < op->bytes) {
      std::cout << "Not enough bytes left to decode " << op->name << 'n';
      it = end;
      return 0; // instruction not found or off the end of the passed array
      }
      return op->decode(it);
      }


      Now main becomes very simple:



      int main(){
      const std::vector<char> char_vect = {0x76, 0x10, 0x20, 0x10}; // Fill vector of char
      auto end{char_vect.cend()};

      for (auto it{char_vect.cbegin()}; it != end; disassembleOne(it, end)) {
      }
      }


      Another way to do this would be to put more of the processing in the Opcode itself -- it would make sense that each opcode would know how to decode itself.



      Be clear about caller expectations



      This code, as with the original code, expects that the it being passed is a valid iterator that is not the end. It is good to document that in comments in the code.






      share|improve this answer











      $endgroup$



      I have some ideas about how you might be able to improve your program.



      Avoid problems



      Rather than trying to deal with the problem for every instruction, one approach is avoiding it entirely. One way to do that is to simply append a number of bytes to the end of the vector. If the maximum bytes for an instruction is $n$, then append $n-1$ bytes to the end and stop when you've advanced into the padded area.



      Check before advancing



      One could also pass the number of remaining bytes to the disassemble function. However, the mechanism I'd suggest would be to pass a range, e.g.:



      int diassembleOne(std::vector<char>::iterator& it, std::vector<char>::iterator& end) {
      // figure out number of bytes for this opcode
      if (std::distance(it, end) > opbytes) {
      it = end;
      // throw or return 0
      }
      // disassemble the instruction thoroughly
      std::advance(it, opbytes);
      return opbytes;
      }


      Use const iterators



      If all the code is doing is disassembling, then it shouldn't alter the underlying vector. For that reason, I'd recommend passing a std::vector<char>::const_iterator &.



      Use classes



      I'd suggest using an Opcode class like this:



      class Opcode {
      char code;
      short int bytes;
      std::string_view name;
      bool operator==(char opcode) const { return code == opcode; }
      int decode(std::vector<char>::const_iterator& it, std::ostream& out=std::cout) const {
      out << name;
      ++it;
      for (int i{bytes-1}; i; --i) {
      out << static_cast<unsigned>(*it++);
      }
      out << 'n';
      return bytes;
      }
      };

      constexpr std::array<Opcode,2> instructions {
      { 0x10, 2, "STOP $" },
      { 0x76, 2, "HALT " },
      };


      Pass a pair of iterators to the dissemble function



      As mentioned before, you can pass a pair of iterators to the disassemble function. Using that plus the class above:



      int disassembleOne(std::vector<char>::const_iterator& it, std::vector<char>::const_iterator& end){
      auto op{std::find(instructions.begin(), instructions.end(), *it)};
      if (op == instructions.end()) {
      std::cout << "Instruction not foundn";
      it = end;
      return 0; // instruction not found or off the end of the passed array
      }
      if (std::distance(it, end) < op->bytes) {
      std::cout << "Not enough bytes left to decode " << op->name << 'n';
      it = end;
      return 0; // instruction not found or off the end of the passed array
      }
      return op->decode(it);
      }


      Now main becomes very simple:



      int main(){
      const std::vector<char> char_vect = {0x76, 0x10, 0x20, 0x10}; // Fill vector of char
      auto end{char_vect.cend()};

      for (auto it{char_vect.cbegin()}; it != end; disassembleOne(it, end)) {
      }
      }


      Another way to do this would be to put more of the processing in the Opcode itself -- it would make sense that each opcode would know how to decode itself.



      Be clear about caller expectations



      This code, as with the original code, expects that the it being passed is a valid iterator that is not the end. It is good to document that in comments in the code.







      share|improve this answer














      share|improve this answer



      share|improve this answer








      edited 12 hours ago

























      answered 12 hours ago









      EdwardEdward

      47.3k378212




      47.3k378212












      • $begingroup$
        The loop in main doesn't use the result of the call to disassembleOne.
        $endgroup$
        – Roland Illig
        9 hours ago










      • $begingroup$
        @RolandIllig It doesn't need to use the return value because the iterator, which is passed by reference, is updated within the function.
        $endgroup$
        – Edward
        8 hours ago


















      • $begingroup$
        The loop in main doesn't use the result of the call to disassembleOne.
        $endgroup$
        – Roland Illig
        9 hours ago










      • $begingroup$
        @RolandIllig It doesn't need to use the return value because the iterator, which is passed by reference, is updated within the function.
        $endgroup$
        – Edward
        8 hours ago
















      $begingroup$
      The loop in main doesn't use the result of the call to disassembleOne.
      $endgroup$
      – Roland Illig
      9 hours ago




      $begingroup$
      The loop in main doesn't use the result of the call to disassembleOne.
      $endgroup$
      – Roland Illig
      9 hours ago












      $begingroup$
      @RolandIllig It doesn't need to use the return value because the iterator, which is passed by reference, is updated within the function.
      $endgroup$
      – Edward
      8 hours ago




      $begingroup$
      @RolandIllig It doesn't need to use the return value because the iterator, which is passed by reference, is updated within the function.
      $endgroup$
      – Edward
      8 hours ago













      4












      $begingroup$

      I think the big switch is a problem. Consider a more data-driven approach where each opcode is described by a struct:



      struct OpCode
      {
      unsigned char command_byte;
      unsigned char mask; // if only a subset of bits determine command
      unsigned char length;
      // other members as needed - e.g. a pointer to the "print" function
      };


      Now, the code that reads instructions can determine whether the command is unterminated, without needing to repeat the logic for every multi-byte opcode.



      I've included the mask so that we don't encode every single instruction (e.g. ld R1, R2 encodes the source and destination registers in the bits of a single-byte command; it would be tedious and error-prone to write them all separately here).



      The simple length value I've shown isn't quite enough, given that the Game Boy's LR35902 processor supports the 0xCB prefix for extended Z80 instructions - we might want to handle that outside of the normal flow, and use it to switch between different instruction tables.

      We get away with a simple length value here, because the only prefix instruction supported by the Game Boy's LR35902 processor is 0xCB, which is always followed by a single-byte instruction. If we were decoding Z80 instructions (with ED prefix), then we'd need something a little more sophisticated.






      share|improve this answer









      $endgroup$


















        4












        $begingroup$

        I think the big switch is a problem. Consider a more data-driven approach where each opcode is described by a struct:



        struct OpCode
        {
        unsigned char command_byte;
        unsigned char mask; // if only a subset of bits determine command
        unsigned char length;
        // other members as needed - e.g. a pointer to the "print" function
        };


        Now, the code that reads instructions can determine whether the command is unterminated, without needing to repeat the logic for every multi-byte opcode.



        I've included the mask so that we don't encode every single instruction (e.g. ld R1, R2 encodes the source and destination registers in the bits of a single-byte command; it would be tedious and error-prone to write them all separately here).



        The simple length value I've shown isn't quite enough, given that the Game Boy's LR35902 processor supports the 0xCB prefix for extended Z80 instructions - we might want to handle that outside of the normal flow, and use it to switch between different instruction tables.

        We get away with a simple length value here, because the only prefix instruction supported by the Game Boy's LR35902 processor is 0xCB, which is always followed by a single-byte instruction. If we were decoding Z80 instructions (with ED prefix), then we'd need something a little more sophisticated.






        share|improve this answer









        $endgroup$
















          4












          4








          4





          $begingroup$

          I think the big switch is a problem. Consider a more data-driven approach where each opcode is described by a struct:



          struct OpCode
          {
          unsigned char command_byte;
          unsigned char mask; // if only a subset of bits determine command
          unsigned char length;
          // other members as needed - e.g. a pointer to the "print" function
          };


          Now, the code that reads instructions can determine whether the command is unterminated, without needing to repeat the logic for every multi-byte opcode.



          I've included the mask so that we don't encode every single instruction (e.g. ld R1, R2 encodes the source and destination registers in the bits of a single-byte command; it would be tedious and error-prone to write them all separately here).



          The simple length value I've shown isn't quite enough, given that the Game Boy's LR35902 processor supports the 0xCB prefix for extended Z80 instructions - we might want to handle that outside of the normal flow, and use it to switch between different instruction tables.

          We get away with a simple length value here, because the only prefix instruction supported by the Game Boy's LR35902 processor is 0xCB, which is always followed by a single-byte instruction. If we were decoding Z80 instructions (with ED prefix), then we'd need something a little more sophisticated.






          share|improve this answer









          $endgroup$



          I think the big switch is a problem. Consider a more data-driven approach where each opcode is described by a struct:



          struct OpCode
          {
          unsigned char command_byte;
          unsigned char mask; // if only a subset of bits determine command
          unsigned char length;
          // other members as needed - e.g. a pointer to the "print" function
          };


          Now, the code that reads instructions can determine whether the command is unterminated, without needing to repeat the logic for every multi-byte opcode.



          I've included the mask so that we don't encode every single instruction (e.g. ld R1, R2 encodes the source and destination registers in the bits of a single-byte command; it would be tedious and error-prone to write them all separately here).



          The simple length value I've shown isn't quite enough, given that the Game Boy's LR35902 processor supports the 0xCB prefix for extended Z80 instructions - we might want to handle that outside of the normal flow, and use it to switch between different instruction tables.

          We get away with a simple length value here, because the only prefix instruction supported by the Game Boy's LR35902 processor is 0xCB, which is always followed by a single-byte instruction. If we were decoding Z80 instructions (with ED prefix), then we'd need something a little more sophisticated.







          share|improve this answer












          share|improve this answer



          share|improve this answer










          answered 11 hours ago









          Toby SpeightToby Speight

          26.2k742118




          26.2k742118























              2












              $begingroup$

              Your code's division into functions isn't very natural. The objective is to have each function performing one specific task, which should also be as orthogonal as possible to the tasks performed by the other functions. For instance, your disassemble function performs three different functions: it reads from the instruction stream, it interprets assembly code and returns the number of bytes that should be skipped to get to the next instruction. That's a not-so-coherent mix of responsibilities.



              There's also a bug in your code, because it+=disassemble(it); could point beyond char_vect.end() which is in itself undefined behavior, even if you don't dereference the iterator.



              I would build my disassembler around an iterator, or better even a range (if you don't mind external libraries or anticipating the next standard):



              #include <algorithm>
              #include <vector>
              #include <iostream>
              #include <range/v3/view.hpp>

              using iterator = std::vector<char>::const_iterator;
              struct truncated_instruction {};

              // one function to interpret the instruction
              void interpret(char instruction) {
              if (instruction == 'x' throw bad_instruction();
              std::cout << (int) instruction << ' ';
              }

              // one function to get the number of bytes to skip
              int nb_bytes(char c) { return 1; }

              class disassembled
              : public ranges::view_facade<disassembled>
              {
              friend ranges::range_access;
              iterator first, last;
              char const & read() const { return *first; }
              bool equal(ranges::default_sentinel) const { return first == last; }
              void next() {
              // one function to get to the next instruction
              auto bytes_to_skip = nb_bytes(*first);
              if (std::distance(first, last) < bytes_to_skip) throw truncated_instruction();
              // check if there's enough space left to advance before advancing
              std::advance(first, bytes_to_skip);
              }
              public:
              disassembled() = default;
              explicit disassembled(const std::vector<char>& v) : first(v.begin()), last(v.end()) {}
              };

              int main() {
              std::vector<char> char_vect = {0x76, 0x10, 0x20, 0x30};
              try {
              for (auto instruction : disassembled(char_vect)) interpret(instruction);
              }
              catch // ...
              }





              share|improve this answer









              $endgroup$


















                2












                $begingroup$

                Your code's division into functions isn't very natural. The objective is to have each function performing one specific task, which should also be as orthogonal as possible to the tasks performed by the other functions. For instance, your disassemble function performs three different functions: it reads from the instruction stream, it interprets assembly code and returns the number of bytes that should be skipped to get to the next instruction. That's a not-so-coherent mix of responsibilities.



                There's also a bug in your code, because it+=disassemble(it); could point beyond char_vect.end() which is in itself undefined behavior, even if you don't dereference the iterator.



                I would build my disassembler around an iterator, or better even a range (if you don't mind external libraries or anticipating the next standard):



                #include <algorithm>
                #include <vector>
                #include <iostream>
                #include <range/v3/view.hpp>

                using iterator = std::vector<char>::const_iterator;
                struct truncated_instruction {};

                // one function to interpret the instruction
                void interpret(char instruction) {
                if (instruction == 'x' throw bad_instruction();
                std::cout << (int) instruction << ' ';
                }

                // one function to get the number of bytes to skip
                int nb_bytes(char c) { return 1; }

                class disassembled
                : public ranges::view_facade<disassembled>
                {
                friend ranges::range_access;
                iterator first, last;
                char const & read() const { return *first; }
                bool equal(ranges::default_sentinel) const { return first == last; }
                void next() {
                // one function to get to the next instruction
                auto bytes_to_skip = nb_bytes(*first);
                if (std::distance(first, last) < bytes_to_skip) throw truncated_instruction();
                // check if there's enough space left to advance before advancing
                std::advance(first, bytes_to_skip);
                }
                public:
                disassembled() = default;
                explicit disassembled(const std::vector<char>& v) : first(v.begin()), last(v.end()) {}
                };

                int main() {
                std::vector<char> char_vect = {0x76, 0x10, 0x20, 0x30};
                try {
                for (auto instruction : disassembled(char_vect)) interpret(instruction);
                }
                catch // ...
                }





                share|improve this answer









                $endgroup$
















                  2












                  2








                  2





                  $begingroup$

                  Your code's division into functions isn't very natural. The objective is to have each function performing one specific task, which should also be as orthogonal as possible to the tasks performed by the other functions. For instance, your disassemble function performs three different functions: it reads from the instruction stream, it interprets assembly code and returns the number of bytes that should be skipped to get to the next instruction. That's a not-so-coherent mix of responsibilities.



                  There's also a bug in your code, because it+=disassemble(it); could point beyond char_vect.end() which is in itself undefined behavior, even if you don't dereference the iterator.



                  I would build my disassembler around an iterator, or better even a range (if you don't mind external libraries or anticipating the next standard):



                  #include <algorithm>
                  #include <vector>
                  #include <iostream>
                  #include <range/v3/view.hpp>

                  using iterator = std::vector<char>::const_iterator;
                  struct truncated_instruction {};

                  // one function to interpret the instruction
                  void interpret(char instruction) {
                  if (instruction == 'x' throw bad_instruction();
                  std::cout << (int) instruction << ' ';
                  }

                  // one function to get the number of bytes to skip
                  int nb_bytes(char c) { return 1; }

                  class disassembled
                  : public ranges::view_facade<disassembled>
                  {
                  friend ranges::range_access;
                  iterator first, last;
                  char const & read() const { return *first; }
                  bool equal(ranges::default_sentinel) const { return first == last; }
                  void next() {
                  // one function to get to the next instruction
                  auto bytes_to_skip = nb_bytes(*first);
                  if (std::distance(first, last) < bytes_to_skip) throw truncated_instruction();
                  // check if there's enough space left to advance before advancing
                  std::advance(first, bytes_to_skip);
                  }
                  public:
                  disassembled() = default;
                  explicit disassembled(const std::vector<char>& v) : first(v.begin()), last(v.end()) {}
                  };

                  int main() {
                  std::vector<char> char_vect = {0x76, 0x10, 0x20, 0x30};
                  try {
                  for (auto instruction : disassembled(char_vect)) interpret(instruction);
                  }
                  catch // ...
                  }





                  share|improve this answer









                  $endgroup$



                  Your code's division into functions isn't very natural. The objective is to have each function performing one specific task, which should also be as orthogonal as possible to the tasks performed by the other functions. For instance, your disassemble function performs three different functions: it reads from the instruction stream, it interprets assembly code and returns the number of bytes that should be skipped to get to the next instruction. That's a not-so-coherent mix of responsibilities.



                  There's also a bug in your code, because it+=disassemble(it); could point beyond char_vect.end() which is in itself undefined behavior, even if you don't dereference the iterator.



                  I would build my disassembler around an iterator, or better even a range (if you don't mind external libraries or anticipating the next standard):



                  #include <algorithm>
                  #include <vector>
                  #include <iostream>
                  #include <range/v3/view.hpp>

                  using iterator = std::vector<char>::const_iterator;
                  struct truncated_instruction {};

                  // one function to interpret the instruction
                  void interpret(char instruction) {
                  if (instruction == 'x' throw bad_instruction();
                  std::cout << (int) instruction << ' ';
                  }

                  // one function to get the number of bytes to skip
                  int nb_bytes(char c) { return 1; }

                  class disassembled
                  : public ranges::view_facade<disassembled>
                  {
                  friend ranges::range_access;
                  iterator first, last;
                  char const & read() const { return *first; }
                  bool equal(ranges::default_sentinel) const { return first == last; }
                  void next() {
                  // one function to get to the next instruction
                  auto bytes_to_skip = nb_bytes(*first);
                  if (std::distance(first, last) < bytes_to_skip) throw truncated_instruction();
                  // check if there's enough space left to advance before advancing
                  std::advance(first, bytes_to_skip);
                  }
                  public:
                  disassembled() = default;
                  explicit disassembled(const std::vector<char>& v) : first(v.begin()), last(v.end()) {}
                  };

                  int main() {
                  std::vector<char> char_vect = {0x76, 0x10, 0x20, 0x30};
                  try {
                  for (auto instruction : disassembled(char_vect)) interpret(instruction);
                  }
                  catch // ...
                  }






                  share|improve this answer












                  share|improve this answer



                  share|improve this answer










                  answered 12 hours ago









                  papagagapapagaga

                  4,547321




                  4,547321






















                      Mai Kar is a new contributor. Be nice, and check out our Code of Conduct.










                      draft saved

                      draft discarded


















                      Mai Kar is a new contributor. Be nice, and check out our Code of Conduct.













                      Mai Kar is a new contributor. Be nice, and check out our Code of Conduct.












                      Mai Kar is a new contributor. Be nice, and check out our Code of Conduct.
















                      Thanks for contributing an answer to Code Review Stack Exchange!


                      • Please be sure to answer the question. Provide details and share your research!

                      But avoid



                      • Asking for help, clarification, or responding to other answers.

                      • Making statements based on opinion; back them up with references or personal experience.


                      Use MathJax to format equations. MathJax reference.


                      To learn more, see our tips on writing great answers.




                      draft saved


                      draft discarded














                      StackExchange.ready(
                      function () {
                      StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fcodereview.stackexchange.com%2fquestions%2f215349%2fdecoding-assembly-instructions-in-a-game-boy-disassembler%23new-answer', 'question_page');
                      }
                      );

                      Post as a guest















                      Required, but never shown





















































                      Required, but never shown














                      Required, but never shown












                      Required, but never shown







                      Required, but never shown

































                      Required, but never shown














                      Required, but never shown












                      Required, but never shown







                      Required, but never shown







                      Popular posts from this blog

                      How did Captain America manage to do this?

                      迪纳利

                      南乌拉尔铁路局