Re: [PATCH v3 1/2] Fix nonnull errors reported by UBSAN with GCC 7.

[Junio C Hamano <gitster@pobox.com>]

Martin Liška <mliska@suse.cz> writes:

> From 0bdf4d717d3d368dd9676d15d20f8592c4d22fde Mon Sep 17 00:00:00 2001
> From: marxin <mliska@suse.cz>
> Date: Wed, 5 Apr 2017 14:31:32 +0200
> Subject: [PATCH 1/2] Fix nonnull errors reported by UBSAN with GCC 7.
>
> Replace call to memmove with newly introduced function memmove_or_null
> and call to memcpy with COPY_ARRAY macro.

I didn't closely follow the discussion, but with these three lines
(which will be the primary thing future readers of this change in
"git log -p" output will rely on), it is unclear why this change was
made.  For that matter, it is not clear what "nonnull errors" are,
either.

> Signed-off-by: Martin Liska <mliska@suse.cz>
> ---
>  apply.c            | 4 +---
>  builtin/ls-files.c | 2 +-
>  git-compat-util.h  | 8 ++++++++
>  3 files changed, 10 insertions(+), 4 deletions(-)
>
> diff --git a/apply.c b/apply.c
> index e6dbab26a..121f3f414 100644
> --- a/apply.c
> +++ b/apply.c
> @@ -2802,9 +2802,7 @@ static void update_image(struct apply_state *state,
>  			img->line + applied_pos + preimage_limit,
>  			(img->nr - (applied_pos + preimage_limit)) *
>  			sizeof(*img->line));
> -	memcpy(img->line + applied_pos,
> -	       postimage->line,
> -	       postimage->nr * sizeof(*img->line));
> +	COPY_ARRAY(img->line + applied_pos, postimage->line, postimage->nr);

I am suspecting that postimage->nr can be 0 and newer compliers can
give warning "hey what's the point of copying 0 bytes?" which can be
squelched by moving to COPY_ARRAY()?  If that is the case I like the
change (but as I said, it is unclear if that is what is going on
here).

> diff --git a/builtin/ls-files.c b/builtin/ls-files.c
> index d449e46db..0a6cc1e8a 100644
> --- a/builtin/ls-files.c
> +++ b/builtin/ls-files.c
> @@ -391,7 +391,7 @@ static void prune_cache(const char *prefix, size_t prefixlen)
>  		}
>  		last = next;
>  	}
> -	memmove(active_cache, active_cache + pos,
> +	memmove_or_null(active_cache, active_cache + pos,
>  		(last - pos) * sizeof(struct cache_entry *));

Does this change come with the same or a similar motivation as the
above (i.e. pos could be the same as last)?

"Something or NULL" is a name we use for a function that returns
something (under normal circumstances) or returns NULL.  This
wrapper is not about returning NULL at all, as far as I can see, and
is misnamed.  If it is about "avoid moving 0 bytes", similar to how
COPY_ARRAY() is used in the previous hunk, perhaps MOVE_ARRAY() is a
better name?

Thanks.

>  	active_nr = last - pos;
>  }
> diff --git a/git-compat-util.h b/git-compat-util.h
> index 8a4a3f85e..81f6e56ac 100644
> --- a/git-compat-util.h
> +++ b/git-compat-util.h
> @@ -1002,6 +1002,14 @@ int git_qsort_s(void *base, size_t nmemb, size_t size,
>  		die("BUG: qsort_s() failed");			\
>  } while (0)
>  
> +static inline void *memmove_or_null(void *dest, const void *src, size_t n)
> +{
> +	if (n > 0)
> +		return memmove(dest, src, n);
> +	else
> +		return dest;
> +}
> +
>  #ifndef REG_STARTEND
>  #error "Git requires REG_STARTEND support. Compile with NO_REGEX=NeedsStartEnd"
>  #endif