How do I save the git-diff from a makefile

Question

I have a project that I compile using a makefile. I want to save the output of git diff into the executable when it is compiled. Simply assigning the output of git diff to a variable in the makefile, and then passing to the compiler e.g.

GIT_DIFF := $(shell git diff)

CXXFLAGS += -D$(GIT_DIFF)

fails because the parts of the string gets evaluated, for example newlines split the compile command into several non-sensical commands.

What is the best way to pass in the output of git diff at compile-time into a string I can use in C++ code?

question from:https://stackoverflow.com/questions/65911886/how-do-i-save-the-git-diff-from-a-makefile

Answer 1

I found a way to do this. Not too sure how robust or portable it is though. This solution assumes the shell is bash. In the makefile I have

GIT_DIFF := $(shell printf '%q
' "$$(git diff | awk '{ gsub(""","\"",$$$$1) ; print $$$$1 }' ORS='\n')" )

CXXFLAGS += -DGIT_DIFF="$(GIT_DIFF)"

and then in a header file I have

// Convert macro to a string constant
#define STRINGIFY1(x) #x
#define STRINGIFY(x) STRINGIFY1(x)

constexpr auto git_diff = STRINGIFY(GIT_DIFF)

(this is a C++11 project).

Unpacking that command a bit:

• $$ escapes the $ symbol in the makefile, so that bash sees the command

  printf '%q
  ' "$(git diff | awk '{ gsub(""","\"",$$1) ; print $$1 }' ORS='\n')"
  

  (I learned this here https://til.hashrocket.com/posts/k3kjqxtppx-escape-dollar-sign-on-makefiles)
• in the awk command the gsub() bit escapes quote marks, because otherwise the C++ compiler (maybe also bash) would get confused. (from here https://www.cyberciti.biz/faq/awk-find-and-replace-fields-values/)
• the ORS='\n' bit of the awk command replaces newlines with \n so that they don't get interpreted as line endings by printf (credit to this blog post https://timmurphy.org/2014/02/12/replacing-newline-characters-in-linux/)
• The printf '%q ' bit uses the bash builtin printf to escape all special characters (this is the bit I know of that may be less portable, needs checking if you want to use a different shell). (Copied from here https://unix.stackexchange.com/a/141323)

The string that's passed to the C++ code is full of escape sequences, so is not very readable. It can be converted back to something more readable by reversing the awk substition and using echo -e to evaluate the escaped characters (for some reason this needs to be done twice). So if the string is saved into a text file, say mygitdiff, then

echo -e $(echo -e "$(cat mygitdiff | awk '{ gsub("\"", """,$$1) ; print $$1 }' IRS='\n')")

should make it something readable.

This method is awkward, and the string that's passed to the C++ isn't very readable, so if anyone has a better solution, or suggestions for improvements, that would be very welcome!