Werror=format-security issue from gprofng/src/Print.cc
Yichao Yu via Gdb <[email protected]>
| Newsgroups | gmane.comp.gdb.devel |
|---|---|
| Message-ID | <CAMvDr+St65VrE2-_boON-m_Ph0A4MhPWFUGqWd9xZ1HQ-z+oxg@mail.gmail.com> |
I got a format-security werror on gprofng/src/Print.cc when trying to build the master version of gdb (the compiler flag is added by the archlinuxcn build machine). While I could disable the flag, I think there might be a real issue looking at the code. The line that causes the issue is https://github.com/bminor/binutils-gdb/blob/master/gprofng/src/Print.cc#L2616, which uses a "dynamic" format string without any argument. AFAICT, the fmt3 is only ever initialized in `er_print_experiment::overview_summary` and if I read it correctly, it's initialized to a string with no actual formatting inputs other than a `%%`. It is used, however, twice in `er_print_experiment::overview_value`, one given two zeros as the arguments and one given no arguments so it looks a bit suspicious. The git log shows now history of this file so I'm not sure what's the intention but my best guess is 1. the `fprintf (out_file, fmt3, 0., 0.);` was meant to be using `fmt4`. (I assume this is to avoid nan from total_value = 0), or 2. since fmt3 is actually a string that's more or less "0.0 (0.0)" with padding, the two `0.`s passed to fmt3 are probably bogus and it should be the same as the `fprintf (out_file, fmt3);` below if my understanding is correct, I think in either case one can simply avoid using fmt3 with fprintf by just removing the extra % from it and directly write it to the output instead. (i.e. https://gist.github.com/yuyichao/7e7cc2f240a1a6e92a1b2a9da8eb3905) Did I miss anything? Yichao