On Sun, 2012-02-12 at 19:34 +0200, Eli Zaretskii wrote: > > Date: Sun, 12 Feb 2012 18:11:49 +0100 > > From: Sebastian Pipping <sebast...@pipping.org> > > CC: psm...@gnu.org, bug-make@gnu.org > > > > Since we would run into buffer overflows with sprintf/vsprintf, we > > rely on snprintf/vsnprintf for that task. Quoting from my "man 3 > > printf" output: > > > > snprintf(), vsnprintf(): > > _BSD_SOURCE || _XOPEN_SOURCE >= 500 || _ISOC99_SOURCE \ > > || _POSIX_C_SOURCE >= 200112L; > > or cc -std=c99 > > > > That's my understanding of the situation. > > If all you need is vsnprintf, I think you can rely on it being > present, and if it isn't, disable argument reordering on that > platform. Or we could import one from libiberty or from gnulib, I > guess. Paul?
I don't see any good way to support systems without vsnprintf(), if we decide to depend on it, except for writing a lot of yucky ifdef'd code scattered throughout the codebase. If we have vsnprintf() we can abstract away the construction of the string into a general function. If we don't, we have to do it down at the very lowest level and pass va_lists around, which means the function signature of the lowest-level functions will have to be different, etc. Pulling in an implementation from gnulib, just in case, might be the way to go. _______________________________________________ Bug-make mailing list Bug-make@gnu.org https://lists.gnu.org/mailman/listinfo/bug-make
