Craig St. Jean's Blog

WebSphere 8 on Windows as Standard User

2011-09-10T17:22:00.005-04:00

While I doubt this is supported usage, I had a goal of using MyEclipse Bling (MyEclipse Blue combined with MyEclipse for Spring), along with IBM WebSphere Application Server v8 on Windows 7 as a standard user.

Here's what I did:

Install WebSphere 8 and MyEclipse Bling

Grant users full control of the AppServer directory

Create a WebSphere profile however you prefer

Open the command prompt as an admin to create a service for the profile

'cd' to your WAS home bin (e.x. cd C:\IBM\WebSphere\AppServer\bin)

WASService.exe -add AppSrv01 -serverName server1 -profilePath C:\IBM\WebSphere\AppServer\profiles\AppSrv01 -startType manual (replace AppSrv01 and the profile path with yours of course)

Download SetACL

Open Services, right click your WAS service and click properties. In Path to executable copy the actual name of the service (e.x. "IBMWAS80Service - AppSrv01")

In the command prompt: SetACL.exe -on "IBMWAS80Service - AppSrv01" -ot srv -actn ace -ace "n:S-1-1-0;p:full;s:y;i:np;m:grant;w:dacl"

The n: part you define the SID of the user you want to grant access. S-1-1-0 is the SID for everyone.

At this point, WAS starts just fine when a standard user starts it via services management, however by using WASService -start (which MyEclipse Blue uses) it still does not work. Update soon when I figure it out.

Compiling g++ for C++0x on CentOS

2011-06-11T09:19:00.014-04:00

CentOS 5 (I don't recall the minor version I am using) includes g++ 4.1, however you cannot do C++0x development without g++ 4.4 or higher. Here's what I did:

Dependencies

First download the dependencies gmp-5.0.2, mpfr-3.0.1, mpc-0.9, ppl-0.11.2, cloog-ppl-0.15.11 (see http://gcc.gnu.org/install/prerequisites.html)

gmp

./configure --enable-cxx
make
make check
sudo make install

mpfr

./configure --with-gmp=/usr/local
make
make check
sudo make install

mpc

./configure --with-gmp=/usr/local
make
make check
sudo make install

ppl

./configure --with-gmp=/usr/local
make
make check
sudo make install

Grab some coffe, food, go for a stroll, etc. during make, this one takes a while (or at least it does in a VM)

cloog-ppl

./configure --with-gmp=/usr/local --with-ppl=/usr/local
make
make check
sudo make install

gcc

Download gcc (I used gcc-4.5.3), gcc-core, and g++

Then from the PARENT directory of where you extracted gcc:

mkdir gccbuild
cd gccbuild
export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
../gcc-4.5.3/configure --enable-languages=c,c++ --with-gmp=/usr/local --with-mpfr=/usr/local --with-mpc=/usr/local --with-ppl=/usr/local --with-cloog=/usr/local
make -j 2
# go to sleep
make check
make install

Local WebSphere Message Broker Maintenance on Windows Domain

2011-06-06T15:03:00.002-04:00

As a follow-up to http://craigstjean.blogspot.com/2011/03/local-websphere-message-broker-install.html, here are some maintenance tips:

After changing your domain password:

Change your local Windows user password to be the same
Open the ODBC Control Panel and update your user in the System DSN
Open the Broker Command Console and type:
mqsichangebroker -i -a -p

Getting Started with WebSphere MQ 7 Broker 7 on Linux

2011-03-12T20:22:00.008-05:00

I was interested in evaluating what it was like working with MQ7 and Broker 7, so I setup a Linux VM (CentOS 5, x64). My goal was to run MQ, Broker, and the Toolkit all locally.

Note: Not all of this is perfect or maybe even ideal, its just what I did. CentOS is not a supported distribution for these products. I know more about MQ6 and Broker 6.1 than 7.

Before I Started

Installed DB2 Express
yum install compat-libstdc++-33 (MQ needs this)

Installing MQ7

Download MQ7 Trial and extract it

./mqlicense.sh -accept (after reading it of course)

rpm -ivh MQSeriesRuntime-7.0.1-3.x86_64.rpm MQSeriesServer-7.0.1-3.x86_64.rpm

rpm -ivh MQSeriesClient-7.0.1-3.x86_64.rpm

rpm -ivh MQSeriesSDK-7.0.1-3.x86_64.rpm

rpm -ivh MQSeriesJava-7.0.1-3.x86_64.rpm

rpm -ivh MQSeriesEclipseSDK33-7.0.1-3.x86_64.rpm

Update /opt/mqm/eclipseSDK33/eclipse/eclipse.ini add -Dorg.eclipse.swt.browser.XULRunnerPath=/usr/lib/firefox-3.6

rpm -ivh MQSeriesConfig-7.0.1-3.x86_64.rpm

usermod -a -G mqm craig (replace craig with your username)

usermod -a -G mqm root

usermod -a -G mqm mqm (for some reason the mqm user didn't have the group)

Installing Message Broker

Download and extract the WMB Trial

groupadd mqbrkrs

usermod -a -G mqbrkrs craig

usermod -a -G mqbrkrs root

./setuplinuxx64 -console

Mostly just enter 1s or 0s until complete

Installing WMB Toolkit

Download and extract the WMB Toolkit Trial

In the IBMInstallationManager directory...

./install -nosplash --launcher.suppressErrors -input mbtoolkit-silent.xml -silent

Create Broker

Since I'm going to do things the easy way instead of the right way (I'm doing everything as root instead of running under mqm), need to reboot here

. /opt/ibm/mqsi/7.0/bin/mqsiprofile (notice the ". " at start)

cd /opt/mqm/bin

./crtmqm -u DLQ ESBBKL

./strmqm ESBBKL

./runmqlsr -m ESBBKL -t TCP &

mqsicreatebroker ESBBKLBKR -i craig -a <password> -q ESBBKL

mqsichangebroker ESBBKLBKR -i craig -a <password> -P 1414

mqsistart ESBBKLBKR

mqsichangeproperties ESBBKLBKR -b httplistener -o HTTPListener -n startListener -v true

mqsistop ESBBKLBKR

mqsistart ESBBKLBKR

mqsicreateexecutiongroup ESBBKLBKR -e default

Working in the Toolkit

In the Brokers view, connect to remote queue ESBBKL at localhost.localdomain port 1414

I know that ideally I could right click Brokers and create a new local broker instead of doing a lot of the above, but obviously something up there threw it off. Furthermore, you can't use localhost because it tells you to use a local broker.

NOTE: When I went through and did this again to verify I didn't miss any steps, the toolkit already had my local broker in it. You probably won't need to add it, however if it is not in the list, do the above.

Create Local Queues

Create file with the following contents

DEFINE QLOCAL ('REQ.MB.TEST_FLOW') +
     REPLACE

DEFINE QLOCAL ('RESP.MB.TEST_FLOW') +
     REPLACE

DEFINE QLOCAL ('DLQ') +
     REPLACE

runmqsc ESBBKL < queuedefs.conf

Create flows, more queues, have fun!

Local WebSphere Message Broker Install on Windows Domain

2011-03-08T12:29:00.004-05:00

After having a lot of difficulty setting up a local WMQ and WMB install on my domain account, here is what I did:

Install MQ and its fixpacks

Install WMB runtime and its fixpacks

Install the Broker Toolkit

Created a local Windows user with the same username/password as my domain user

Add user to mqbrkrs and mqm groups

Install latest DB2 Express

Start the broker command console

MQSI 6.1.0.5
C:\IBM\MQSI\6.1
Start broker Command Console

C:\IBM\MQSI\6.1>cd c:\IBM\WMQ\bin

c:\IBM\WMQ\bin>crtmqm.exe -u DLQ ESBBKL
WebSphere MQ queue manager created.
Creating or replacing default objects for ESBBKL.
Default objects statistics : 43 created. 0 replaced. 0 failed.
Completing setup.
Setup completed.

c:\IBM\WMQ\bin>strmqm ESBBKL
WebSphere MQ queue manager 'ESBBKL' starting.
5 log records accessed on queue manager 'ESBBKL' during the log replay phase.
Log replay for queue manager 'ESBBKL' complete.
Transaction manager state recovered for queue manager 'ESBBKL'.
WebSphere MQ queue manager 'ESBBKL' started.

c:\IBM\WMQ\bin>call mqsicreatedb ESBBKLDB -e DB2
BIP9809I: Creating DB2 Database 'ESBBKLDB', this may take a while.
BIP9825I: Creating the ODBC Data Source name 'ESBBKLDB'.
BIP9817I: The DB2 database engine was used to create the 'ESBBKLDB' database.

BIP8071I: Successful command completion.

c:\IBM\WMQ\bin>call mqsicreateconfigmgr ESBBKL_CFGMGR -i <USERNAME> -a <PASSWORD> -q ESBBKL
AMQ8110: WebSphere MQ queue manager already exists.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
BIP8071I: Successful command completion.

c:\IBM\WMQ\bin>call mqsicreatebroker ESBBKLBKR -i <USERNAME> -a <PASSWORD> -q ESBBKL -n ESBBKLDB
AMQ8110: WebSphere MQ queue manager already exis
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
The setmqaut command completed successfully.
BIP8071I: Successful command completion.

c:\IBM\WMQ\bin>call mqsistart ESBBKL_CFGMGR
BIP8096I: Successful command initiation, check the system log to ensure that the component started without problem and that it continues to run without problem.

c:\IBM\WMQ\bin>call mqsistart ESBBKLBKR
BIP8096I: Successful command initiation, check the system log to ensure that the component started without problem and that it continues to run without problem.

Open MQ Explorer

Create new queue named DLQ in ESBBKL

Create new TCP Listener (I used port 1414, the default)

In the toolkit, add new Domain Connection to ESBBKL port 1414 on localhost

Connect to the new Domain

Add new broker reference to the topology, ESBBKLBKR on ESBBKL

Enjoy!

Note: if you want to delete and start over, do NOT delete the database first. If you do, call mqsicreatedb again, then use mqsichangebroker and set your username and password again.

The steps I follow to delete are:

Start broker command console and cd c:\IBM\WMQ\bin

call mqsistop ESBBKLBKR

call mqsideletebroker ESBBKLBKR

call mqsistop ESBBKL_CFGMGR

call mqsideleteconfigmgr -w -q ESBBKL_CFGMGR

call mqsideletedb -d DeleteAllDatabases

Seattle's Best: French Roast Coffee

2009-05-31T08:49:00.001-04:00

Buy it
Drink it
Repeat

Problems installing DB2 on Windows Vista

2009-04-01T19:24:00.005-04:00

Installed DB2 Express on Windows Vista (Business 64-bit) yesterday, and hit a bunch of errors at the end of the install.

Specifically:

An error occurred while initializing the Profile Registry.

The return value is "-2079391743". Contact your technical service
representative.

An error occurred while setting the registry key "SOFTWARE\Microsoft
\MSDTC\XADLL\C:\Program Files\IBM\SQLLIB\BIN\DB2APP.DLL" to "C:
\Program Files\IBM\SQLLIB\BIN\DB2APP.DLL".

One or more errors occurred while committing the changes to the
user "DB2". Create or make any changes to this user manually.

An error occurred while creating the instance "DB2". The return
code is "-1918". Create the instance manually using the command
"db2icrt".

So after some research, here is how to fix it:

Create DAS user (for example db2admin) via Windows as an Administrator

Execute "db2admin CREATE /USER:db2admin /PASSWORD:*password*"

Execute "db2admin SETID db2admin"

Execute "db2icrt -u db2admin DB2" (where DB2 is the instance name)

Set System Environment Variable "DB2INSTANCE" to your instance name

Set System Environment Variable "DB2PATH" to "C:\Program Files\IBM\SQLLIB"

Reboot your machine

Follow http://publib.boulder.ibm.com/infocenter/db2luw/v9r5/topic/com.ibm.db2.luw.qb.server.doc/doc/t0011930.html to add yourself to DB2ADMNS

Execute "db2set db2comm=tcpip"

Start your DB2 instance

Firebird on Glassfish

2009-03-29T20:54:00.003-04:00

Ok, so after a few hours of "pulling my hair out" (I don't actually have any...), I've finally managed to get Glassfish to work with Firebird.

Configuration:
Firebird 2.1.1(.17910-0_Win32 Classic)
Jaybird JDBC Driver 2.1.6 (JDK 1.6)
Sun GlassFish Enterprise Server 2.1

How-to:
0. Stop your domain
1. Copy jaybird-2.1.6.jar and jaybird-pool-2.1.6.jar into your domain's lib
Do NOT put the JARs into ext like I did. It ended up being the biggest source of pain in trying to get this to work.
2. Start your domain
3. Log into the admin console
4. Select Resources->JDBC->Connection Pools -> New
5. Enter a name and select javax.sql.ConnectionPoolDataSource resource type (leave vendor blank) and click Next
6. Enter 'org.firebirdsql.pool.sun.AppServerConnectionPoolDataSource' for Datasource Classname
7. Enter the following additional properties (note, I am using UTF-8, you may need to change your charSet and encoding (or just omit them))
charSet = UTF-8
encoding = UNICODE_FSS
userName = XXX
password = XXX
database = localhost:C:/Databases/DATABASE.FDB
databaseName = localhost:C:/Databases/DATABASE.FDB
(Yes, databaseName is deprecated in Jaybird, however the connection pool datasource won't work without it)
8. Enjoy

Fun with assembler (on FreeBSD)

2009-01-19T23:40:00.002-05:00

Searching is the heart of artificial intelligence, but what do you do when the core of your algorithms is much larger than it should be? Optimize!

Take this code for example

#include <assert.h>
#include <string.h>

#define NUM_NODES 10
#define MAX_STACK 10

int adjMatrix[NUM_NODES][NUM_NODES];
int visited[NUM_NODES];

int stack[MAX_STACK];
int stack_index;

void initAdjMatrix(void)
{
 memset((void *)adjMatrix, 0, sizeof(adjMatrix));
 memset((void *)visited, 0, sizeof(visited));
}

void makeEdge(int from, int to)
{
 assert(from < NUM_NODES);
 assert(to < NUM_NODES);

 adjMatrix[from][to] = 1;
}

void initStack(void)
{
 stack_index = 0;
}

void push(int elem)
{
 assert(stack_index < MAX_STACK);
 stack[stack_index++] = elem;
}

int pop(void)
{
 int elem;
 assert(stack_index > 0);
 elem = stack[--stack_index];
 return elem;
}

int empty(void)
{
 if (stack_index == 0)
  return 1;
 return 0;
}

void dfs(int start, int goal)
{
 int node, i;

 push(start);
 while (!empty())
 {
  node = pop();

  if (node == goal)
  {
   printf("%d Goal\n", node);
   return;
  }

  if (visited[node])
   continue;
  else
   visited[node] = 1;

  printf("%d ", node);

  for (i = NUM_NODES - 1; i > -1; i--)
   if (adjMatrix[node][i] == 1)
    push(i);
 }

 printf("Goal not found!\n");
}

int main(int argc, char *argv[])
{
 initAdjMatrix();
 initStack();

 //make our graph
 makeEdge(1, 2);
 makeEdge(1, 3);
 makeEdge(1, 4);
 makeEdge(2, 5);
 makeEdge(3, 5);
 makeEdge(3, 6);
 makeEdge(4, 7);
 makeEdge(5, 8);
 makeEdge(5, 9);

 dfs(1, 6);

 return 0;
}

(Slightly modified version of the depth first search from "AI Application Programming" Second Edition by M. Tim Jones)

Lets run that through gcc:


[craig ~/Desktop/ai/dfs]$ gcc -o dfs dfs.c
[craig ~/Desktop/ai/dfs]$ ls -l dfs
-rwxr-xr-x  1 craig  wheel  6029 Aug 18 22:53 dfs

6029 bytes!!!

Ok, lets tell gcc to optimize it:


[craig ~/Desktop/ai/dfs]$ gcc -O3 dfs.c -o dfs
[craig ~/Desktop/ai/dfs]$ ls -l dfs
-rwxr-xr-x  1 craig  wheel  6274 Aug 18 22:54 dfs

Horray for sacraficing space for speed... I want both though!

Lets take a look at the code sections just out of curiosity


Sections:
Idx Name          Size      VMA       LMA       File off  Algn
  0 .interp       00000015  080480d4  080480d4  000000d4  2**0
                  CONTENTS, ALLOC, LOAD, READONLY, DATA
  1 .hash         00000098  080480ec  080480ec  000000ec  2**2
                  CONTENTS, ALLOC, LOAD, READONLY, DATA
  2 .dynsym       00000130  08048184  08048184  00000184  2**2
                  CONTENTS, ALLOC, LOAD, READONLY, DATA
  3 .dynstr       000000cb  080482b4  080482b4  000002b4  2**0
                  CONTENTS, ALLOC, LOAD, READONLY, DATA
  4 .rel.plt      00000030  08048380  08048380  00000380  2**2
                  CONTENTS, ALLOC, LOAD, READONLY, DATA
  5 .init         00000011  080483b0  080483b0  000003b0  2**2
                  CONTENTS, ALLOC, LOAD, READONLY, CODE
  6 .plt          00000070  080483c4  080483c4  000003c4  2**2
                  CONTENTS, ALLOC, LOAD, READONLY, CODE
  7 .text         00000428  08048434  08048434  00000434  2**2
                  CONTENTS, ALLOC, LOAD, READONLY, CODE
  8 .fini         0000000c  0804885c  0804885c  0000085c  2**2
                  CONTENTS, ALLOC, LOAD, READONLY, CODE
  9 .rodata       00000113  08048868  08048868  00000868  2**0
                  CONTENTS, ALLOC, LOAD, READONLY, DATA
 10 .data         0000000c  0804997c  0804997c  0000097c  2**2
                  CONTENTS, ALLOC, LOAD, DATA
 11 .eh_frame     00000004  08049988  08049988  00000988  2**2
                  CONTENTS, ALLOC, LOAD, READONLY, DATA
 12 .dynamic      00000098  0804998c  0804998c  0000098c  2**2
                  CONTENTS, ALLOC, LOAD, DATA
 13 .ctors        00000008  08049a24  08049a24  00000a24  2**2
                  CONTENTS, ALLOC, LOAD, DATA
 14 .dtors        00000008  08049a2c  08049a2c  00000a2c  2**2
                  CONTENTS, ALLOC, LOAD, DATA
 15 .jcr          00000004  08049a34  08049a34  00000a34  2**2
                  CONTENTS, ALLOC, LOAD, DATA
 16 .got          00000024  08049a38  08049a38  00000a38  2**2
                  CONTENTS, ALLOC, LOAD, DATA
 17 .bss          0000022c  08049a60  08049a60  00000a60  2**5
                  ALLOC
 18 .comment      0000012d  00000000  00000000  00000a60  2**0
                  CONTENTS, READONLY

hmm...that seems unnecessary

Time for a rewrite!


bits 32

; Define our constants
%define stdout  1
%define SYS_exit 1
%define SYS_write 4
%define NUM_NODES_SIZE_1 10
%define NUM_NODES 100
%define MAX_STACK 10

; Define our messages
segment .data
goalMsg  db 'Goal', 0x0A
goalMsgBytes equ $ - goalMsg

spaceMsg db ' '
spaceMsgBytes equ $ - spaceMsg

goalNotFoundMsg db 'Goal not found', 0x0A
goalNotFoundMsgBytes equ $ - goalNotFoundMsg

numbers db '0123456789'
output db 0, ' '

; Declare our arrays
segment .bss
adjMatrix resb NUM_NODES
visited  resb MAX_STACK

segment .text

align 4
; ==============
; = kernelCall =
; ==============
kernelCall:
 int 0x80
 ret

; Define our macros
%macro system  1
 mov eax, %1
 call kernelCall
%endmacro

%macro sys.exit  0
 system SYS_exit
%endmacro

%macro sys.write 0
 system SYS_write
%endmacro

; ===========================================
; = initAdjMatrix                           =
; = Parameters: none                        =
; = Pre: none                               =
; = Post: adjMatrix and visited are cleared =
; ===========================================
align 4
initAdjMatrix:
 push eax

 ; Clear adjMatrix
 ; Clear all of eax because we reference it later
 xor eax, eax
.initAdjMatrixL1:
 mov [adjMatrix + eax], byte 0
 inc al
 cmp al, NUM_NODES
 jne .initAdjMatrixL1

 ; Clear visited
 xor al, al
.initAdjMatrixL2
 mov [visited + eax], byte 0
 inc al
 cmp al, MAX_STACK
 jne .initAdjMatrixL2

 pop eax
 ret

; ==============================================
; = makeEdge                                   =
; = Parameters: from node (ebx), to node (ecx) =
; = Pre: none                                  =
; = Post: matrix[from][to] is set              =
; ==============================================
align 4
makeEdge:
 mov ebx, [esp + 8]
 mov ecx, [esp + 4]
 ; Determine row (from * 10)
 ; Plus column (to)
 imul ebx, ebx, NUM_NODES_SIZE_1
 add ebx, ecx
 mov [adjMatrix + ebx], byte 1

 ret

; =============================================
; = dfs                                       =
; = Parameters: start (eax), goal (edx)       =
; = Pre: Stack and matrix must be initialized =
; = Post: Graph is traversed                  =
; =============================================
align 4
dfs:
 push ebp
 mov ebp, esp

 ; Set start
 mov eax, [esp + 12]
 ; Set goal
 mov edx, [esp + 8]
 
 ; Make room for stack_index
 sub esp, 2

 ; Using ebp - 2 as stack_index
 mov [ebp - 2], word 0
 
 ; Push start and increase stack index
 push eax
 inc word [ebp - 2]
.dfsL1:
 ; Loop until stack_index is 0
 cmp [ebp - 2], word 0
 jne .dfsL12
 jmp .dfsOutNotFound
.dfsL12:
 ; Retrieve node
 pop eax
 dec word [ebp - 2]

 ; Check if node is our goal
 cmp eax, edx
 je .dfsOutFound

 ; If we've visited node, continue
 cmp [visited + eax], byte 0
 jne .dfsL1
 ; Set visited to 1
 mov [visited + eax], byte 1

 ; Output eax
 push eax
 mov al, [numbers + eax]
 mov [output], al
 push dword 2
 push dword output
 push dword stdout
 sys.write
 add esp, byte 12
 pop eax

 ; Loop from 9 to 0
 xor ecx, ecx
 mov cl, NUM_NODES_SIZE_1
 dec cl
.dfsL2:
 cmp cl, -1
 je .dfsL1

 ; Calculate place in array
 push eax
 imul eax, eax, NUM_NODES_SIZE_1
 add eax, ecx
 ; Compare our adjMatrix to 1
 cmp [adjMatrix + eax], byte 1
 ; If not 1, loop
 jne .dfsL3
 pop eax
 ; Otherwise push our index
 push ecx
 inc word [ebp - 2]
 dec cl
 jmp .dfsL2
.dfsL3:
 pop eax
 dec cl
 jmp .dfsL2

.dfsOutFound:
 ; Output eax
 push eax
 mov al, [numbers + eax]
 mov [output], al
 push dword 2
 push dword output
 push dword stdout
 sys.write
 add esp, byte 12
 pop eax
 
 ; Output goal message
 push dword goalMsgBytes
 push dword goalMsg
 push dword stdout
 sys.write
 add esp, byte 12

 jmp .dfsLeave
.dfsOutNotFound:
 ; Output goal not found message
 push dword goalNotFoundMsgBytes
 push dword goalNotFoundMsg
 push dword stdout
 sys.write
 add esp, byte 12
.dfsLeave:
 mov esp, ebp
 pop ebp
 ret

; ========
; = main =
; ========
global _start
_start:
 call initAdjMatrix

 ; Make our edges
 push 1
 push 2
 call makeEdge
 push 1
 push 3
 call makeEdge
 push 1
 push 4
 call makeEdge
 push 2
 push 5
 call makeEdge
 push 3
 push 5
 call makeEdge
 push 3
 push 6
 call makeEdge
 push 4
 push 7
 call makeEdge
 push 5
 push 8
 call makeEdge
 push 5
 push 9
 call makeEdge
 
 ; Traverse
 push 1
 push 6
 call dfs

 push dword 0
 sys.exit

I've already made a few optimizations (for example changing things like "mov eax, 0" to "xor eax, eax", or had I needed it "mov eax, 1" would become "xor eax, eax" "inc eax". Yes its 2 statements but it is actually 2 bytes smaller - longer code does not always mean longer machine code)

And lets build it:


[craig ~/Desktop/ai/dfs/asm]$ nasm -f elf -l dfs.lst dfs.asm
[craig ~/Desktop/ai/dfs/asm]$ ld -s -o dfs dfs.o
[craig ~/Desktop/ai/dfs/asm]$ ls -l dfs
-rwxr-xr-x  1 craig  wheel  964 Aug 18 22:57 dfs

Ok ok we are MUCH better, but we can do a little bit better. Let's take a look at the sections:


[craig ~/Desktop/ai/dfs/asm]$ objdump -x dfs
...SNIP...
Sections:
Idx Name          Size      VMA       LMA       File off  Algn
  0 .text         000001eb  08048080  08048080  00000080  2**4
                  CONTENTS, ALLOC, LOAD, READONLY, CODE
  1 .data         00000021  0804926c  0804926c  0000026c  2**2
                  CONTENTS, ALLOC, LOAD, DATA
  2 .bss          00000070  08049290  08049290  00000290  2**2
                  ALLOC
  3 .comment      0000001f  00000000  00000000  00000290  2**0
                  CONTENTS, READONLY
...SNIP...

Hmm, my guess is we really don't need the .comment section.


[craig ~/Desktop/ai/dfs/asm]$ hexdump -C -s 0x290 -n 31 dfs
00000290  00 54 68 65 20 4e 65 74  77 69 64 65 20 41 73 73  |.The Netwide Ass|
000002a0  65 6d 62 6c 65 72 20 30  2e 39 38 2e 33 39 00     |embler 0.98.39.|
000002af

Looks like nasm is making us pay for room to setup the .comment section, and is making us pay for a little recognition -- lets fix that.


[craig ~/Desktop/ai/dfs/asm]$ strip -R .comment dfs
[craig ~/Desktop/ai/dfs/asm]$ ls -l dfs
-rwxr-xr-x  1 craig  wheel  884 Aug 18 23:04 dfs

884 bytes! We could do better by creating a binary file and defining the elf headers ourselves, but we've already surpassed the goal of the optimizations - 6.82 times smaller!

And if anyones curious:


[craig ~/Desktop/ai/dfs]$ ls -l dfs.c
-rw-r--r--  1 craig  wheel  1373 Aug 18 21:11 dfs.c
[craig ~/Desktop/ai/dfs]$ ls -l asm/dfs.asm
-rw-r--r--  1 craig  wheel  4310 Aug 18 21:55 asm/dfs.asm
[craig ~/Desktop/ai/dfs]$ wc -l dfs.c
     103 dfs.c
[craig ~/Desktop/ai/dfs]$ wc -l asm/dfs.asm
     260 asm/dfs.asm