Re: Covierty Integration / Improvement

[Johannes Schindelin <Johannes.Schindelin@gmx.de>]

Hi Theodore,

On Wed, 6 Apr 2022, Theodore Ts'o wrote:

> On Wed, Apr 06, 2022 at 05:08:37PM +0200, Johannes Schindelin wrote:
> > I have fixed Git for Windows' Coverity build and started to sift through
> > the 154 new defects reported as of v2.36.0-rc0.
> >
> > Sadly, there is now a new class of overwhelming false positives: Coverity
> > claims that "strbuf_addstr does not [NUL-]terminate", which is of course
> > false.
>
> It should be possible to suppress this by uploading a Coverity model
> file.  See[1] for more details:
>
> [1] https://community.synopsys.com/s/article/practical-example-of-coverity-function-model

Right, I know about this, my apologies for being too succinct. For the
record, here is how we submit the Coverity builds in Git for Windows:

https://github.com/git-for-windows/build-extra/blob/bea0c37bdc3737843b7808269b077b30e51c67fb/please.sh#L1793-L1925

Unfortunately, you cannot see the model file because that is not provided
per build, but globally for the entire project. Here is the model file we
currently use:

-- snip --
/* modelfile for git */

char strbuf_slopbuf[64];

void *malloc(size_t);
void *calloc(size_t, size_t);
void *realloc(void *, size_t);
void free(void *);

void *xrealloc(void *ptr, size_t size)
{
	void *ret = realloc(ptr, size);
	if (!ret) __coverity_panic__();
	return ret;
}

void *xmalloc(size_t size)
{
	void *mem = malloc(size);
	if (!mem) __coverity_panic__();
	return mem;
}

void xcalloc(size_t num, size_t size)
{
	void *ret = calloc(num, size);
	if (!ret)  __coverity_panic__();
	return ret;
}

void usage(const char *err) {
  __coverity_panic__();
}

void usagef(const char *err, ...) {
  __coverity_panic__();
}

void die(const char *err, ...)  {
  __coverity_panic__();
}

void die_errno(const char *err, ...) {
  __coverity_panic__();
}
-- snap --

I _guess_ we could "help" Coverity by providing some alternative
implementation of `strbuf_add()` that does not use `memcpy()` but instead
a loop that explicitly NUL-terminates the string.

But it feels wrong to do that because that would weaken the analysis
because Coverity would not actually analyze the code that is executed
anymore.

> I've suppressed a similar issue by using the attribute __nonstring,
> but I don't think that will work for git, because strbuf->buf really
> *is* a NUL-terminated string, where as in ext4 we have some fields
> which are designed to be NUL padded, but it is *not* guaranteed to be
> NUL-terminated:
>
> #ifndef __nonstring
> #ifdef __has_attribute
> #if __has_attribute(__nonstring__)
> #define __nonstring                    __attribute__((__nonstring__))
> #else
> #define __nonstring
> #endif /* __has_attribute(__nonstring__) */
> #else
> # define __nonstring
> #endif /* __has_attribute */
> #endif /* __nonstring */
>
> struct ext2_super_block {
>        ...
> /*068*/	__u8	s_uuid[16] __nonstring;		/* 128-bit uuid for volume */
> /*078*/	__u8	s_volume_name[EXT2_LABEL_LEN] __nonstring;	/* volume name */
>        ...
> };
>
> (This is needed to suppress warnings by Clang as well.)
>
> Using __nonstring will result in attempts to use s_volume_name in "C"
> string context to give a warning, which is why this isn't right for
> strbuf->buf.

Correct. I guess we could fool around with the model file until those
false positives are gone, but I have to admit that I cannot justify the
time to work on this.

Ciao,
Johannes