Re: Lint for Makefiles

[Kaz Kylheku (gmake)]

On 2021-04-02 00:55, Christian Hujer wrote:
> Hello everyone.
>
> I was looking for a Lint-like tool for Makefiles.
> The ones I could find either seemed abandoned (like mint) or had a
> broken build (checkmake) and other fundamental flaws (not operating on
> included Makefiles).
> I hope to get some responses on this list here for more tools that
> people have created.
> Because I couldn't find one that works for me, I created one.
>
> The first two checks implemented are, which isn't much but at least a 
> start:
> • Report direct usage of `rm` (should use `$(RM)` instead).

Note that POSIX doesn't specify any RM variable.

Look for the "Default Rules" section here

https://pubs.opengroup.org/onlinepubs/9699919799/utilities/make.html

> This one
> was the reason for writing this tool, a dev had used rm instead of
> $(RM) in clean, causing it to fail on consecutive runs.

The actual bug was not using "rm -f". I see that RM gives you that:

  $ make -p | grep '^RM ='
  RM = rm -f

Since this is nonstandard though, it would behoove you to define it
yourself, or at least weakly with ?=

  RM ?= rm -f

Unless you are sure that your Makefile is strictly GNU Make.

Aliases for common utilities that are quite portable, and are not
toolchain components don't make a whole lot of sense, IMHO.

If we are cross-compiling, the build machine's "rm" should work just
fine for removing something in a build directory, or even target
sysroot; we have no reason to be putting a cross toolchain prefix
onto rm or anything like that.

If it was the case that some versions of "rm" don't have the "-f"
option (and it is implied), yet other versions of "rm" require
it, then the abstraction would be valuable.

But "rm" and "rm -f" are standardized by POSIX, like make itself.
A Makefile that uses "rm -f ..." in a recipe line can be
entirely POSIX conforming.