Re: [Cvs-dev] Re: Result of CVS Coverity scan, via NetBSD

[Christos Zoulas]

On May 12,  6:35am, address@hidden ("Mark D. Baushke") wrote:
-- Subject: Re: [Cvs-dev] Re: Result of CVS Coverity scan, via NetBSD

| -----BEGIN PGP SIGNED MESSAGE-----
| Hash: SHA1
| 
| Hi Christos,
| 
| Would it make sense to modify src/gnu/usr.bin/cvs/include/config.h
| so that it does a
| 
| /* Functions defined in src/gnu/dist/xcvs/src/subr.c */
| #define xasprintf cvs_xasprintf
| #define xmalloc   cvs_xmalloc
| #define xrealloc  cvs_xrealloc
| #define xstrdup   cvs_xstrdup
| 
| or, does the coverity model processing not handle macro processing
| correctly?

It does, and it will work properly [provided it compiles :-)]

| For what it is worth, I have not finished going through all of the run
| 22 reported problems. I know that there is at least a few more that are
| 'real' bugs (cid-1058 and cid-1061 are both such).
| 
| Given that our xstrdup() is safe, I will probably be considering reverting
| the=20
| 
|     var =3D str ? xstrdup (str) : NULL;
| 
| changes back into
| 
|     var =3D xstrdup (str);

Yes, we should revert the changes.
 
| in the CVS STABLE and CVS FEATURE source bases. (It turns out that CVS
| FEATURE uses Xstrdup to do the NULL check before calling the real
| xstrdup () function, but has a '#define xstrdup Xstrdup' to hide it.)

Ouch, pre-processor macro hell...


christos