Stuart Henderson: > Perhaps by calling a different function instead of v*printf that does > most of the same things except not syslogging for null %s prints..
Well... glib2's configure checks for vasprintf() and vsnprintf() and can fall back to an included copy of gnulib's implementation of those functions. If I understand the gnulib code correctly, it doesn't actually reimplement the *printf() family, but works as a wrapper that breaks down some functionality and still calls basic system *printf() functions. I also can't find any special handling of NULL arguments for the %s format. So I think forcing glib to use the gnulib implementation will not solve this problem by itself. However, the gnulib functions call an internal printf_parse() routine that analyzes the format string and digests it into arrays of directives and corresponding arguments. This looks like the spot where an argument for %s could be checked for NULL and swapped out with a pointer to the string constant "(null)". Assuming we manage to hack up the g_log() backend as outlined above, what's next? With the most common source of those log messages removed, the second most common one will become apparent... I question our wisdom of having printf() log instances of %s NULL, but then declaring that we can't fix the offenders and need to resort to intercepting arguments passed to printf(). -- Christian "naddy" Weisgerber na...@mips.inka.de
