Bug 29675
| Summary: | lint fails on stdio.h | ||
|---|---|---|---|
| Product: | Base System | Reporter: | Walter Campbell <wcampbel> |
| Component: | bin | Assignee: | freebsd-bugs (Nobody) <bugs> |
| Status: | Closed FIXED | ||
| Severity: | Affects Only Me | ||
| Priority: | Normal | ||
| Version: | 4.3-STABLE | ||
| Hardware: | Any | ||
| OS: | Any | ||
>>Number: 29675 >>Category: bin >>Synopsis: lint fails on stdio.h >>Confidential: no >>Severity: non-critical >>Priority: low >>Responsible: freebsd-bugs >>State: open >>Quarter: >>Keywords: >>Date-Required: >>Class: sw-bug >>Submitter-Id: current-users >>Arrival-Date: Mon Aug 13 08:20:01 PDT 2001 >>Closed-Date: >>Last-Modified: >>Originator: Walter Campbell >>Release: FreeBSD 4.3-STABLE i386 >>Organization: >botbay.net >>Environment: >System: FreeBSD botbay.net 4.3-STABLE FreeBSD 4.3-STABLE #0: Fri Jul 27 >11:53:35 EDT 2001 root@botbay.net:/usr/obj/usr/src/sys/KABEL i386 >Also checked with current cvsup (as of 11h03 EDT) >>Description: >lint(1) will fail on C files that include stdio.h AFAIK, or rather, AFAICROTL (As Far As I Can Read On The Lists), lint(1) has not been [actively] mantained for quite a while. You may wish to search the archives for this topic. AFAICS, instead of lint(1), gcc options are used. You maight want to take a look at the Warning options in gcc(1). HTH (<gasp> yet another acronym :-)), Salvo On Mon, 13 Aug 2001, Walter Campbell wrote: > >Description: > > lint(1) will fail on C files that include stdio.h > > >How-To-Repeat: > > wcampbel@botbay (alis-2.2beta2.patched): lint -aacgprxhH *.c > alis.c: > stdio.h:299: syntax error > ... > This happens with make lint on ircd-hybrid 6.2rc1 > ... What version of lint is in that? I think thre is no problem here with FreeBSD lint, but stdio.h is certainly broken for most foreign versions of lint, compilers other than gcc, and old version of gcc. The hard-coded __attribute__(()) in it needs to be ifdef'ed, not quite as in the following patch. --- Index: stdio.h =================================================================== RCS file: /home/ncvs/src/include/stdio.h,v retrieving revision 1.34 diff -u -2 -r1.34 stdio.h --- stdio.h 2001/04/22 01:56:08 1.34 +++ stdio.h 2001/05/04 02:28:31 @@ -69,7 +69,4 @@ }; -/* hold a buncha junk that would grow the ABI */ -struct __sFILEX; - /* * stdio state variables. @@ -116,5 +113,5 @@ /* separate buffer for long sequences of ungetc() */ struct __sbuf _ub; /* ungetc buffer */ - struct __sFILEX *_extra; /* additions to FILE to not break ABI */ + struct __sFILEX *_extra; /* additions to FILE to not break ABI */ int _ur; /* saved _r when _r is counting ungetc data */ @@ -263,7 +260,4 @@ FILE *fdopen __P((int, const char *)); int fileno __P((FILE *)); -int ftrylockfile __P((FILE *)); -void flockfile __P((FILE *)); -void funlockfile __P((FILE *)); __END_DECLS #endif /* not ANSI */ @@ -300,10 +294,16 @@ int asprintf __P((char **, const char *, ...)) __printflike(2, 3); char *ctermid_r __P((char *)); -__const char *fmtcheck __P((const char *, const char *)) - __attribute__((__format_arg__(2))); char *fgetln __P((FILE *, size_t *)); +void flockfile __P((FILE *)); +__const char *fmtcheck __P((const char *, const char *)) +#ifdef __GNUC__ + __attribute__((__format_arg__(2))) +#endif + ; int fpurge __P((FILE *)); int fseeko __P((FILE *, _BSD_OFF_T_, int)); _BSD_OFF_T_ ftello __P((FILE *)); +int ftrylockfile __P((FILE *)); +void funlockfile __P((FILE *)); int getw __P((FILE *)); int pclose __P((FILE *)); @@ -349,7 +349,7 @@ __BEGIN_DECLS int __srget __P((FILE *)); -int __vfscanf __P((FILE *, const char *, _BSD_VA_LIST_)); int __svfscanf __P((FILE *, const char *, _BSD_VA_LIST_)); int __swbuf __P((int, FILE *)); +int __vfscanf __P((FILE *, const char *, _BSD_VA_LIST_)); __END_DECLS --- Don't use the other parts of this patch. A more complicated ifdef like the ones in <sys/cdef.h> is required to support old versions of gcc. Bruce On Tue, 14 Aug 2001, Bruce Evans wrote: > On Mon, 13 Aug 2001, Walter Campbell wrote: > > > >Description: > > > > lint(1) will fail on C files that include stdio.h > > > > >How-To-Repeat: > > > > wcampbel@botbay (alis-2.2beta2.patched): lint -aacgprxhH *.c > > alis.c: > > stdio.h:299: syntax error > > ... > > This happens with make lint on ircd-hybrid 6.2rc1 > > ... > > What version of lint is in that? I think thre is no problem here with > FreeBSD lint, but stdio.h is certainly broken for most foreign versions > of lint, compilers other than gcc, and old version of gcc. The > hard-coded __attribute__(()) in it needs to be ifdef'ed, not quite as > in the following patch. I'm using the stock FreeBSD lint. It does give me the syntax error, which causes all my lint's to fail. In response to an earlier reply, lint does things that gcc's -W flags do not, such as identifying unused structures, functions, and helping identify some pieces of code that can be troublesome. On Mon, 13 Aug 2001, Walter Campbell wrote: > On Tue, 14 Aug 2001, Bruce Evans wrote: > > > On Mon, 13 Aug 2001, Walter Campbell wrote: > > > wcampbel@botbay (alis-2.2beta2.patched): lint -aacgprxhH *.c > > > alis.c: > > > stdio.h:299: syntax error > > > ... > > > This happens with make lint on ircd-hybrid 6.2rc1 > > > ... > > > > What version of lint is in that? I think thre is no problem here with > > FreeBSD lint, but stdio.h is certainly broken for most foreign versions > > of lint, compilers other than gcc, and old version of gcc. The > > hard-coded __attribute__(()) in it needs to be ifdef'ed, not quite as > > in the following patch. > > I'm using the stock FreeBSD lint. It does give me the syntax error, > which causes all my lint's to fail. I see. I tested under -current and FreeBSD lint doesn't give the syntax error there. This is a bug in -current. FreeBSD lint now uses "/usr/bin/cc -E x c -D__attribute__(x)= ..." for the preprocessor instead of the correct /usr/bin/cpp (in it 4.3 uses /usr/libexec/cpp which is correct there). Using cc instead of cpp results in __GNUC__ and possibly other wrong things being defined. This would break all programs that carefully use "#ifdef __GNUC ...", except some of the breakage is worked around by defining away __attribute__(x). But defining away __attribute__(x) mainly breaks lint's detection of use of this unportable feature. Bruce State Changed From-To: open->closed Close at request of the committer: "the new lint import fixed it" |
lint(1) will fail on C files that include stdio.h Fix: I am not sure how to fix this. How-To-Repeat: wcampbel@botbay (alis-2.2beta2.patched): lint -aacgprxhH *.c alis.c: stdio.h:299: syntax error alis.c:56: warning: argument id unused in function channel_hash_func alis.c:126: warning: pointer casts may be troublesome alis.c:133: warning: pointer casts may be troublesome alis.c:144: warning: pointer casts may be troublesome alis.c:162: warning: pointer casts may be troublesome alis.c:176: warning: pointer casts may be troublesome alis.c:202: warning: pointer casts may be troublesome alis.c:251: warning: declaration hides earlier one: tmp alis.c:753: warning: statement not reached alis.c:711: warning: argument argc unused in function main alis.c:711: warning: argument argv unused in function main This happens with make lint on ircd-hybrid 6.2rc1 netdb.h:127: warning: struct sockaddr never defined dbuf.c: stdio.h:299: syntax error make lint will then error out with return code 1. Sample code that can cause lint to error out: #include <stdio.h> int main(int argc, char *arvg[]) { printf("Test\n"); return 0; } Giving the messages: wcampbel@botbay (wcampbel): lint -aacgprxhH lint-test.c lint-test.c: stdio.h:299: syntax error lint-test.c:3: warning: argument argc unused in function main lint-test.c:3: warning: argument arvg unused in function main wcampbel@botbay (wcampbel): echo $? 1