1) Besides "-Wformat=2", gcc accepts any option in the form of -Wformat=<number> Based on the manual, gcc only accepts "-Wformat=2", and others such as "-Wformat=0,1, 3-9" should be invalid. Any undocumented behavior? 2) Is the warning name "[-Wformat=]" in the warning message intended? $: gcc-trunk -Wformat=3 s.c s.c: In function ‘main’: s.c:4:3: warning: format ‘%s’ expects argument of type ‘char *’, but argument 2 has type ‘int’ [-Wformat=] printf("%s%s", i, i); ^ s.c:4:3: warning: format ‘%s’ expects argument of type ‘char *’, but argument 3 has type ‘int’ [-Wformat=] $: 3) Is the column number information correct in the warning message? Should the warning refer to the location of the problematic argument, or the function call? Currently clang warns at the argument. $: clang-trunk -Wall s.c s.c:4:18: warning: format specifies type 'char *' but the argument has type 'int' [-Wformat] printf("%s%s", i, i); ~~ ^ %d s.c:4:21: warning: format specifies type 'char *' but the argument has type 'int' [-Wformat] printf("%s%s", i, i); ~~ ^ %d 2 warnings generated. $
(In reply to Chengnian Sun from comment #0) > 1) Besides "-Wformat=2", gcc accepts any option in the form of > -Wformat=<number> > Based on the manual, gcc only accepts "-Wformat=2", and others such as > "-Wformat=0,1, 3-9" should be invalid. Any undocumented behavior? Well, -Wformat=0 disables all format warnings (not really documented) - and should be equivalent to -Wno-format; "-Wformat=1" is equivalent to "-Wformat" - and that is explicitly documented in the manual: http://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#index-Wformat-294 And -Wformat=2 is equivalent to -Wformat=1 plus additional warnings. -Wformat=3 would be equivalent to -Wformat=2 plus additional (nonexisting) options. And I think it currently acts as -Wformat=2, which should be fine as well. (Similar is: -O0, -O1, -O2, -O3; where some other compilers have -O4 - but GCC doesn't and [I think it] treats it as -O3.) > 2) Is the warning name "[-Wformat=]" in the warning message intended? Yes - as one doesn't want to distinguish between "-Wformat=2" and "-Wformat/-Wformat=1 -Wformat-security". In particular, "-Wformat-security" requires "-Wformat=1" in addition, otherwise, it doesn't warn at all. Thus, printing [-Wformat=] is way easier than to re-construct the exact option. > 3) Is the column number information correct in the warning message? Should > the warning refer to the location of the problematic argument, or the > function call? Well, the column number is correct - but pointing to either to the proper '%' in the format string or to the argument would be better. - Like always, there is room for improvement.
Just adding a bit more info: 1) Like Tobias said, it is intended. 2) I agree that reproducing the exact flags that trigger the warning would be useful, but it is a lot of work that nobody so far found time to address. It seems also quite low priority compared to other things. 3) is PR52952, so I am closing this as a duplicate of that. I did some work, but then I got stuck. I don't have enough free time anymore to continue it, so if someone wishes to pick it up, please do so. *** This bug has been marked as a duplicate of bug 52952 ***
Thanks, Tobias and Manuel.