Re: Re: How to transfrom a.c dir1/b.c to -I. -I./dir1

[Philip Guenther]

On Fri, Aug 1, 2008 at 2:34 AM, PRC <address@hidden> wrote:
...
> Did you mean using the function `realpath`? But it is too old to
> support this function for the version of 3.8 of `Make`. Maybe it's more
> simple to invoke shell functions to do this.

'realpath'?  Why would you need that function to do what you want?
The manual describes realpath like this:

`$(realpath NAMES...)'
     For each file name in NAMES return the canonical absolute name.
     ...

But your question had nothing to do with canonicalization:

> There is a variable in my Makefile script, whose value is:
> src = a.c dir1/b.c c.c dir2/d.c
> I wonder how to transform its value to
> -I. -I./dir1 -I. -I./dir2

That sounds like simple text manipulation to me: you want to extract
the directory part of each filename and then prepend "-I" to it, no?
For the 'extract directory part', doesn't the function $(dir) on that
same page seem like an obvious first step for a solution?

$ cat Makefile
src = a.c dir1/b.c c.c dir2/d.c
all:
        @echo $(src)
        @echo $(dir $(src))
$ make
a.c dir1/b.c c.c dir2/d.c
./ dir1/ ./ dir2/
$

Looks like $(dir) leaves a trailing slash, so you need to trim that
and add a leading "-I".  That sort of manipulation is easy with
$(patsubst), described on the previous page in the manual:

$ cat Makefile
src = a.c dir1/b.c c.c dir2/d.c
all:
        @echo $(src)
        @echo $(dir $(src))
        @echo $(patsubst %/,-I%,$(dir $(src)))
$ make
a.c dir1/b.c c.c dir2/d.c
./ dir1/ ./ dir2/
-I. -Idir1 -I. -Idir2
$

Do that process of breaking the goal down into steps and debugging it
with a test makefile make sense?


Philip Guenther