Re: [RFC PATCH] Allow annotating calloc for Valgrind

[Rich Felker <dalias@libc.org>]

On Fri, Jun 30, 2017 at 01:56:14AM +0300, Alexander Monakov wrote:
> With mal0_clear entered only for larger-than-one-page allocations, a
> few additional nops and stack accesses don't matter.
> 
> This only needs Valgrind headers to build, Valgrind itself doesn't need to be
> built/installed.  However, valgrind.h unconditionally includes stdarg.h.

Use of valgrind annotation was already rejected a long time ago. The
same can be done with a suppressions file and that's where it belongs.

> Untested.
> ---
>  configure           | 11 +++++++++++
>  src/malloc/malloc.c |  4 ++++
>  2 files changed, 15 insertions(+)
> 
> diff --git a/configure b/configure
> index c2db298c..7e78094f 100755
> --- a/configure
> +++ b/configure
> @@ -30,6 +30,7 @@ System types:
>  Optional features:
>    --enable-optimize=...   optimize listed components for speed over size [auto]
>    --enable-debug          build with debugging information [disabled]
> +  --enable-valgrind       add Valgrind annotations [disabled]
>    --enable-warnings       build with recommended warnings flags [disabled]
>    --enable-visibility     use global visibility options to optimize PIC [auto]
>    --enable-wrapper=...    build given musl toolchain wrapper [auto]
> @@ -134,6 +135,7 @@ build=
>  target=
>  optimize=auto
>  debug=no
> +valgrind=no
>  warnings=no
>  visibility=auto
>  shared=auto
> @@ -161,6 +163,8 @@ case "$arg" in
>  --disable-optimize) optimize=no ;;
>  --enable-debug|--enable-debug=yes) debug=yes ;;
>  --disable-debug|--enable-debug=no) debug=no ;;
> +--enable-valgrind|--enable-valgrind=yes) valgrind=yes ;;
> +--disable-valgrind|--enable-valgrind=no) valgrind=no ;;
>  --enable-warnings|--enable-warnings=yes) warnings=yes ;;
>  --disable-warnings|--enable-warnings=no) warnings=no ;;
>  --enable-visibility|--enable-visibility=yes) visibility=yes ;;
> @@ -390,6 +394,13 @@ tryflag CFLAGS_MEMOPS -fno-tree-loop-distribute-patterns
>  #
>  test "$debug" = yes && CFLAGS_AUTO=-g
>  
> +if test x"$valgrind" = xyes  ; then
> +CPPFLAGS="$CPPFLAGS -DVG_MEMCHECK_H='<memcheck.h>'"
> +else
> +CPPFLAGS="$CPPFLAGS -DVG_MEMCHECK_H='</dev/null>'"
> +fi
> +CPPFLAGS="${CPPFLAGS# }"
> +
>  #
>  # Preprocess asm files to add extra debugging information if debug is
>  # enabled, our assembler supports the needed directives, and the
> diff --git a/src/malloc/malloc.c b/src/malloc/malloc.c
> index b56fdaa2..d4c229bd 100644
> --- a/src/malloc/malloc.c
> +++ b/src/malloc/malloc.c
> @@ -8,6 +8,7 @@
>  #include "libc.h"
>  #include "atomic.h"
>  #include "pthread_impl.h"
> +#include VG_MEMCHECK_H

This looks really flaky, and makes compilation dependent on a
POSIX-conforming filesystem. For example it almost certainly breaks
cross-compiling from a mingw-based cross compiler (which
musl-cross-make supports building).

The same could be achieved without hacks just using #if/#ifdef around
the #include. But I don't see how <memcheck.h> could work anyway; musl
obviously does not, and can't, use any host include paths where a
header might already be installed.

Rich