zoltan-balazs / jabyinja

My Eötvös Loránd University (Computer Science BSc.) Thesis

Ja(va) By(tecode) In(terpreter) (in) Ja(va)

My BSc thesis work for Eötvös Loránd University's Computer Science curriculum

Prerequisities

• Code:

  • JDK17
  • Maven 3.9.0

• Thesis:

  Arch Linux

  • biber
  • ghostscript
  • pgf-umlcd
  • texlive-bibtexextra
  • texlive-full
  Debian / Ubuntu

  • biber
  • ghostscript
  • texlive-bibtex-extra
  • texlive-font-utils
  • texlive-lang-european
  • texlive-latex-base
  • texlive-pictures
  • texlive-science
  MacOS

  • biber
  • ghostscript
  • texlive
  Other distros

  Find the closest matching packages, or install the whole of latex from the official source

  • Remove existing installation:

    rm -rf /usr/local/texlive/20*
    rm -rf ~/.texlive20*

  • Create a folder for texlive:

    sudo mkdir -p /usr/local/texlive/$(date +%Y)
    sudo chown -R "$USER" /usr/local/texlive

  • Install from the official source:

    mkdir /tmp/texlive
    cd /tmp/texlive
    wget http://mirror.ctan.org/systems/texlive/tlnet/install-tl.zip
    unzip ./install-tl.zip -d install-tl
    cd install-tl/install-tl-$(date +%Y%m%d)
    perl ./install-tl

    Once in perl console, enter i

  • Set variables:

    export MANPATH="$MANPATH:/usr/local/texlive/$(date +%Y)/texmf-dist/doc/man"
    export INFOPATH="$INFOPATH:/usr/local/texlive/$(date +%Y)/texmf-dist/doc/info"
    export PATH=/usr/local/texlive/$(date +%Y)/bin/x86_64-linux:$PATH

  TeX Packages

  Only install if on MacOS / there is an error in generating the pdf

  • adjustbox
  • algpseudocode
  • amsfonts
  • amsmath
  • amsthm
  • array
  • listingsutf8
  • longtable
  • makecell
  • minted
  • multirow
  • paralist
  • pgf-umlcd
  • rotating
  • subcaption
  • todonotes

  $ tlmgr install adjustbox algpseudocode amsfonts \
          amsmath amsthm array listingsutf8 longtable \
          makecell minted multirow paralist pgf-umlcd \
          rotating subcaption todonotes

• Other:

  • Just (optional - for use with justfile)

Building the code

Normally:

$ mvn package
$ java -jar target/jabyinja-*.jar <optional class file>

With Just:

$ just co
$ java -jar target/jabyinja-*.jar <optional class file>

Generating a PDF from the thesis

Ignore any warnings

Normally:

$ cd thesis/
$ pdflatex thesis.tex
$ bibtex thesis
$ pdflatex thesis.tex
$ pdflatex thesis.tex

With Just:

$ just t

Usage

Without specifying any files, the program will try to interpret Main.class, located in the same directory as from where you ran the program from. If you do intend to specifcy a given file, do it with a single argument, where you name the .class file (i.e. if your file is named Example.class, you write Example.class NOT Example)

Number precision

Jabyinja

	BYTE	SHORT	INTEGER	LONG	FLOAT	DOUBLE
MAX	127	32767	2147483647	9223372036854775807	3.4028235E38	1.7976931348623157E308
MIN	-128	-32768	-2147483648	-9223372036854775808	1.4E-45	4.9E-324
0xCA	-54	-54	-54	-54
0xFE	-2	-2	-2	-2
0xBA	-70	-70	-70	-70
0xBE	-66	-66	-66	-66
0xCAFE	-2	-13570	-13570	-13570
0xBABE	-66	-17730	-17730	-17730
0xCAFEBABE	-66	-17730	-889275714	-889275714
0xCAFEBABECAFEBABE	-66	-17730	-889275714	-3819410105351357762
PI (6)					3.141592	3.141592
PI (7)					3.1415925	3.1415926
PI (6) ^ 2					9.8696	9.869600294464002
PI (6) ^ 3					31.006258	31.006257328285756
PI (15)					3.1415927	3.141592653589793
PI (16)					3.1415927	3.141592653589793
PI (15) ^ 2					9.869605	9.869604401089358
PI (15) ^ 3					31.006279	31.006276680299816

/usr/bin/java

	BYTE	SHORT	INTEGER	LONG	FLOAT	DOUBLE
MAX	127	32767	2147483647	9223372036854775807	3.4028235E38	1.7976931348623157E308
MIN	-128	-32768	-2147483648	-9223372036854775808	1.4E-45	4.9E-324
0xCA	-54	-54	-54	-54
0xFE	-2	-2	-2	-2
0xBA	-70	-70	-70	-70
0xBE	-66	-66	-66	-66
0xCAFE	-2	-13570	-13570	-13570
0xBABE	-66	-17730	-17730	-17730
0xCAFEBABE	-66	-17730	-889275714	-889275714
0xCAFEBABECAFEBABE	-66	-17730	-889275714	-3819410105351357762
PI (6)					3.141592	3.141592
PI (7)					3.1415925	3.1415926
PI (6) ^ 2					9.8696	9.869600294464002
PI (6) ^ 3					31.006258	31.006257328285756
PI (15)					3.1415927	3.141592653589793
PI (16)					3.1415927	3.141592653589793
PI (15) ^ 2					9.869605	9.869604401089358
PI (15) ^ 3					31.006279	31.006276680299816

Thankfully, they are the exact same

JVM Instructions checklist

OPCODE	IMPLEMENTED?
NOP
ACONST_NULL
ICONST_*
LCONST_*
FCONST_*
DCONST_*
BIPUSH
SIPUSH
LDC*
ILOAD*
LLOAD*
FLOAD*
DLOAD*
ALOAD*
IALOAD
LALOAD
FALOAD
DALOAD
AALOAD
BALOAD
CALOAD
SALOAD
ISTORE*
LSTORE*
FSTORE*
DSTORE*
ASTORE*
IASTORE
LASTORE
FASTORE
DASTORE
AASTORE
BASTORE
CASTORE
SASTORE
POP*
DUP*
DUP2*
SWAP
*ADD
*SUB
*MUL
*DIV
*REM
*NEG
*SHL
*SHR
*USHR
*AND
*OR
*XOR
IINC
I2*
L2*
F2*
D2*
LCMP
FCMP*
DCMP*
IF*
GOTO
JSR
RET
TABLESWITCH
LOOKUPSWITCH
*RETURN
*STATIC
*FIELD
INVOKE*
INVOKEDYNAMIC
NEW
NEWARRAY
ANEWARRAY
ARRAYLENGTH
ATHROW
CHECKCAST
INSTANCEOF
MONITOR*
WIDE
MULTIANEWARRAY
IF*NULL
GOTO_W
JSR_W

jabyinja's Issues

If the program contains a Double / Long, the class file isn't correctly read

This is due https://docs.oracle.com/javase/specs/jvms/se7/html/jvms-4.html#jvms-4.4.5

A Double / Long entry takes up 2 places in the constant pool as they both take up 8 bytes

Consider using a logging library instead of println's for debug info

Incorrect parsing of file

The current Main.java file is being parsed incorrectly.

A BufferUnderflowException is being triggered, meaning that either readShort or readInt cannot read enough bytes from the buffer.

Having a long/double as argument of a static method breaks it

Due to long/double values being 8 byte wide, they take up 2 physical spaces in the local variables

Make sure all bytecode instructions are correctly implemented

For example, INVOKEs should push their result to the stack (they do not do that currently)

There are only 4 local variables implemented, the specification requires 256 (byte)

Calling any STORE instruction results in an exception

Converting a removed Object from the stack isn't as easy as just casting it. We need to convert it to a Number object first and call the appropriate method for the value.

Start implementing ClassFile parsing

Parse ClassFile as per language specifications: https://docs.oracle.com/javase/specs/jvms/se7/html/jvms-4.html

Calling any CONV instructions throws an exception

Due to not being able to convert Objects as easily, we first need to convert it to appropriate type first, then convert it again.

Long numbers are imprecisely parsed

Due to reading in a signed integer, the low bytes of a long can be incorrect fairly easily.

Double & Float numbers are imprecise

Due to using builtin long's and basic power functions, floating point precision breaks fairly easily.

Refactor ClassFile parsing

Rethinking if the current structure is good.
Is it clean enough?
Is there not too much repetition?

DUP instructions index the stack incorrectly

All DUP instructions have an off-by-one indexing error

DUP/DUP2 instructions can fail if the stack size is low enough

Due to indexing out of the stack an IndexOutOfRange exception can happen.

Implement testing for parsing

To better ensure that the Class file is parsed correctly, testing must be implemented.

• As of the issue's opening, I have not figured out what testing library to use
• I need reproducible cases to check validity of Class file parsing
• Also need some sample files to parse (probably "Hello World", simple List application, etc.)

Start implementing code running

For now a simple "Hello World" programming being interpreted is fine.
Everything is got to start somewhere.