1197205 - [meta] get rid of PR_s{,n}printf et al calls

Status: NEW

(Core :: General, defect)

Description

[Nathan Froyd [:froydnj]]

PR_s{,n}printf has a couple of issues:

- It doesn't accept the same set of formatting characters as the standard s{,n}printf does.
- Some of the formatting characters it does accept mean different things than the standard ones do (e.g. %ld means "32-bit int", not "long int").
- It doesn't support format string checking on compilers that support that (GCC and I think clang).

We'd like to reduce our dependence on NSPR, so this is a good place to start.  Just using s{,n}printf is fairly straightforward; we do have to use the mozilla/Snprintf.h polyfill until we move to MSVC 2015, though.

The only tricky cases are printing pointers (%p) and floats (%f, %g).  Ideally the only time we print either of those with PR_s{,n}printf are in log files or debug mode messages, so web content or internal tests won't depend on the exact formatting.

Let's use this as a meta-bug for splitting up the work amongst directories.

Comment 1

[Tom Tromey :tromey]

After bug 1060419, most calls to PR_snprintf are in security/nss and nspr itself.
However there are still a few probably-dubious uses in libstagefright, of the form

#if _MSC_VER < 1900
#include <nspr/prprf.h>
#define snprintf PR_snprintf
#endif