Build script for an operating system written in C++

Question

I recently started following the James M Kernel Dev tutorials and was able to create an operating system in C++.

Here is the source on GitHub.

I have a couple of questions:

1. Is my coding style ok?
2. Is it wrong to write bash scripts for building the project?

JOSMake.sh

#
# Build File for JOS
#
# To Build
# bash build.sh
# To Run
# bash run.sh
# ----------------------------------------------------------------------------------------------
# JOS uses assembly to build the bootsector
# Compile the boot Sectors in boot
echo "==============================================="
echo "-------- Building Boot Sectors ----------------"
cd boot
# Run the Make Script inside the boot directory
bash buildboot.sh
cd ..
echo "==============================================="
# ----------------------------------------------------------------------------------------------
# ----------------------------------------------------------------------------------------------
# Compile the C++ Sources into Respective Objects
# Compile main.cc
echo "==============================================="
echo "--------- Compiling C++ Sources --------------"
g++ -c -nostdlib -nostdinc -fno-builtin -fno-stack-protector -ffreestanding -m32 -o main.o main.cc
mv main.o build
# Run scons to build the Entire Lbrary of C++ Files
cd include
rm *.o
scons
# Move all Object Files into Build Directory
cp *.o ../build
cd ..
echo "==============================================="
# ----------------------------------------------------------------------------------------------
# ----------------------------------------------------------------------------------------------
## When Everything is Finished Link together the Objects into a Kernel File
echo "==============================================="
echo "--------- Linking ----------------"
cd build
# require ld , link.ld
ld -T '../link.ld' -m elf_i386 -o kernel.jos *.o
# Clean Up After Build
#rm *.o
echo "==============================================="
# ----------------------------------------------------------------------------------------------
# ----------------------------------------------------------------------------------------------
## When Everything is Finished Run the Kernel
qemu-system-i386 -kernel kernel.jos
# -----------------------------------------------------------------------------------------

Answer 1

Instead of repeated statements like this:

echo "==============================================="

I suggest adding a helper function, for example:

print_hr() { # hr - as in horizontal rule
 echo "==============================================="
}

The advantage of this is if you later decide to change the length of the = in the line, you can change it in one place.

---

Instead of entering into sub-directories with cd dirname and later stepping back with cd .., I recommend to use sub-shell environments with (...). For example, instead of this:

cd include
rm *.o
scons
# Move all Object Files into Build Directory
cp *.o ../build
cd ..

Do like this:

(
 cd include
 rm *.o
 scons
 # Move all Object Files into Build Directory
 cp *.o ../build
)

Notice that there is no more cd ... Working directory changes in a (...) sub-shell environment are only in effect within.

---

This maybe a matter of taste, but I find that the many comment lines like this are just noise, and make the script hard to read, I suggest to remove them:

# ----------------------------------------------------------------------------------------------

Answer 2

It's not wrong to use a Bash script to build your project, but it is suboptimal and unconventional. Usually, a makefile is preferred, since Make can compare timestamps and rebuild just the objects that need to be rebuilt.

There are other advantages to makefiles as well. For example, a common convention would be to define variables like

CXX = g++
CXXFLAGS = -nostdlib -nostdinc -fno-builtin -fno-stack-protector -ffreestanding -m32 $(EXTRA_CXXFLAGS)

Then, you would have the flexibility to run make EXTRA_CXXFLAGS=-O2 to compile with -O2 without having to edit the script.

---

A lot of your script's sequences of operations seem complicated and error-prone. For example:

cd include
rm *.o
scons
# Move all Object Files into Build Directory
cp *.o ../build
cd ..

When your script changes directories, that's a bit of global state information that you need to keep track of. That makes the script "brittle" — deleting or rearranging a few lines of the script can have massive effects on all subsequent commands. If you need to change directories, use a subshell as @janos suggested. Better yet, avoid changing directories at all, if possible — and you should be able to build a simple project like this without jumping around.

Why are there .o files in the include directory? By convention, a directory named include would only contain .h header files that are included elsewhere, and don't need to be compiled (unless you are using precompiled headers — but that's not the case here).

Why copy the .o files into the ../build directory? Find some way to get the compiler to produce them in the right place the first time (using the -o option). Copying is inefficient, messes up timestamps, and leaves unnecessary junk.

---

Your script violates the Single Responsibility Principle:

## When Everything is Finished Run the Kernel
qemu-system-i386 -kernel kernel.jos

If you advertise your script as a build script, then it should be just that. Imagine you were packaging this software as a component in a Linux distribution. Having such a line in the build script would be a serious annoyance.

• aswin mohan: Thanks ! I will dump the bash script and Rewrite it in make .