"Richard W.M. Jones" <rjones@redhat.com> wrote:
On Wed, Mar 26, 2008 at 08:55:53PM +0100, Jim Meyering wrote:
qemudReportError(NULL, NULL, NULL, VIR_ERR_NO_MEMORY, - "vncTLSx509certdir"); + "%s", _("failed to allocate vncTLSx509certdir"));
versus:
qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, - "Cannot find QEMU binary %s: %s", binary, + _("Cannot find QEMU binary %s: %s"), binary, strerror(errno));
I assume that the problem with the first one is that gettext might erroneously return a string containing % sequences, resulting in a runtime failure or even exploit. But that could also be a problem with the second one too, surely? (ie. gettext might return three or more % sequences).
Right. But there's nothing we can do about the latter, and there *is* something we can do about the former: precede with "%s". Of course, gettext tools like msgmerge are careful to ensure that %-directives in translations match those in the original, so it's not a problem, in general.
OCaml gettext offers two forms of the gettext function, one for plain strings and one for format strings[1]. The format string version checks that any % sequences in the translated string are compatible with those in the original string. (If not then the original string is returned to avoid any exploit). Sounds as if we need a similar
Nice.
feature in C gettext. A cursory check of the info file didn't show anything like this.
Rich.
[1] and of course the powerful type system ensures that you always use the correct form, ho hum ...
;-)