On Thu, Apr 4, 2013 at 6:49 PM, Ted Unangst <[email protected]> wrote:
> To prevent the future recurrence of some rather serious libc build
> bugs, such as a macro like strong_alias suddenly turning into an
> implicit function prototype, I think we should add some warnings and
> use as much -Werror as possible. (I'm not entirely sure this would
> have caught the missing strong_alias macro, but it seems like a good
> idea anyway.)
>
> This adds a (very) few warnings and fixes some of the fallout. The
> diff is a bit of a mix of things, since I initially tried -Wall before
> I ran away screaming. Over time, maybe some brave pioneers can expand
> the frontier.
...
> +CFLAGS+=-Wimplicit -Wparentheses -Wbounded -Werror
The make release process builds parts of libc with extra #defines to
disable stuff; have you done that with these flags to verify there
aren't warnings in those chunks?
> --- regex/regcomp.c 7 Nov 2011 09:58:27 -0000 1.21
> +++ regex/regcomp.c 5 Apr 2013 01:22:39 -0000
> @@ -171,8 +171,7 @@ regcomp(regex_t *preg, const char *patte
> len = strlen((char *)pattern);
>
> /* do the mallocs early so failure handling is easy */
> - g = (struct re_guts *)malloc(sizeof(struct re_guts) +
> - (NC-1)*sizeof(cat_t));
> + g = (struct re_guts *)malloc(sizeof(struct re_guts));
...
> --- regex/regex2.h 30 Nov 2004 17:04:23 -0000 1.7
> +++ regex/regex2.h 5 Apr 2013 01:22:39 -0000
> @@ -149,7 +149,7 @@ struct re_guts {
> int backrefs; /* does it use back references? */
> sopno nplus; /* how deep does it nest +s? */
> /* catspace must be last */
> - cat_t catspace[1]; /* actually [NC] */
> + cat_t catspace[NC]; /* actually [NC] */
> };
Please tell me that used to actually _use_ the struct hack and didn't
always allocate the full size!
> --- regex/regexec.c 5 Aug 2005 13:03:00 -0000 1.11
> +++ regex/regexec.c 5 Apr 2013 01:22:39 -0000
> @@ -140,6 +140,8 @@ regexec(const regex_t *preg, const char
> regmatch_t pmatch[], int eflags)
> {
> struct re_guts *g = preg->re_g;
> + char *s = (char *)string; /* XXX fucking gcc XXX */
What's the real problem? gcc loses the cast when it inlines smatcher/lmatcher?
> --- rpc/xdr_rec.c 1 Sep 2010 14:43:34 -0000 1.15
> +++ rpc/xdr_rec.c 5 Apr 2013 01:22:39 -0000
> @@ -409,6 +409,7 @@ xdrrec_destroy(XDR *xdrs)
> mem_free(rstrm, sizeof(RECSTREAM));
> }
>
> +bool_t __xdrrec_getrec(XDR *xdrs, enum xprt_stat *statp, bool_t expectdata);
It can't be made static? I see no other references in base; check
xenocara and ask sthen@ to run a grep of the ports source...
> --- stdio/fputws.c 9 Nov 2009 00:18:27 -0000 1.5
> +++ stdio/fputws.c 5 Apr 2013 01:22:39 -0000
> @@ -35,6 +35,8 @@
> #include <wchar.h>
> #include "local.h"
>
> +wint_t __fputwc_unlock(wchar_t wc, FILE *fp);
Why not add that to stdio/local.h?
Philip Guenther
