Re: AC_EGREP_CPP and AC_EGREP_HEADER users - beware of recent gcc

[Mike Frysinger]

On 15 Feb 2017 11:52, Bruno Haible wrote:
> Some uses of AC_EGREP_CPP and AC_EGREP_HEADER probably assume that when
> no continuation lines (backslash-newline) and no multiline macro invocations
> are involved, each source line produces at most one output line.
> 
> This is no longer the case with GCC >= 5.
> 
> Example:
> =============== foo.c ===============
> #include <features.h>
> #define FOO 2
> #define FOO_VERSION 12
> 
> Version FOO . FOO_VERSION
> Version __GLIBC__ . __GLIBC_MINOR__
> 
> #define show(a,b) Version a . b
> show(FOO,FOO_VERSION)
> show(__GLIBC__,__GLIBC_MINOR__)
> =====================================
> 
> With gcc-4.9.4:
> $ gcc -E foo.c
> [...]
> 
> Version 2 . 12
> Version 2 . 23
> 
> 
> Version 2 . 12
> Version 2 . 23
> 
> With gcc-5.1.0 and newer:
> $ gcc -E foo.c
> [...]
> 
> Version 2 . 12
> Version 
> # 6 "foo.c" 3 4
>        2 
> # 6 "foo.c"
>                  . 
> # 6 "foo.c" 3 4
>                    23
> 
> 
> 
> # 9 "foo.c"
> Version 2 . 12
> Version 
> # 10 "foo.c" 3 4
> 2 
> # 10 "foo.c"
> . 
> # 10 "foo.c" 3 4
> 23
> 
> 
> As you can see, here 1 source line produces up to 4 output lines (ignoring the
> line number comment lines). Attempts to use 'grep Version' to find the values 
> of
> __GLIBC__ and __GLIBC_VERSION__ don't work any more.
> 
> Maybe the documentation of AC_EGREP_CPP and AC_EGREP_HEADER could include
> a caveat about this?

seems like we be using -P with CPP when it's supported ?
-mike

signature.asc
Description: Digital signature