On Fri, Sep 04, 2015 at 03:17:31PM +1000, Damien Miller wrote:
> Hi,
>
> Comments appreciated.
as micm@ already mentioned, utf8_stringprep.c in patch is in two (or
three ?) copies of itself in the same file.
> diff --git a/sshconnect2.c b/sshconnect2.c
> index 2b525ac..04120e7 100644
> --- a/sshconnect2.c
> +++ b/sshconnect2.c
> @@ -455,21 +457,51 @@ input_userauth_error(int type, u_int32_t seq, void
> *ctxt)
> return 0;
> }
>
> +/* Check whether we can display UTF-8 safely */
> +static int
> +utf8_ok(void)
> +{
> + static int ret = -1;
> + char *cp;
> +
> + if (ret == -1) {
> + setlocale(LC_CTYPE, "");
> + cp = nl_langinfo(CODESET);
> + ret = strcmp(cp, "UTF-8") == 0;
> + }
> + return ret;
> +}
> +
it could be have a possible problem here. but I don't known enough ssh
code to be sure.
setlocale(1) isn't thread safe, and it changes the behaviour of
multiple functions.
with static variable `ret', you ensure it is called only once. but the
program will don't have the same behaviour in all its life-time:
- between start and the first `utf8_ok' call, it will use C locale,
- and after, it will use the implementation-defined native environment
(value from LC_* and LANG environment variable).
for openbsd, which currently support only C and UTF-8 locales it should
be ok. I don't see others functions than isspace(3) type (isalnum(3),
...) in ssh, and it shouldn't be difference between C and UTF-8 (they
share the same ctype table for 0-255). but for portable, it could be
wrong.
but be aware that some functions, like, `mblen' (used in sftp.c, so not
in same context than `utf8_ok' I think), could copte badly with
setlocale() call (man mblen extract):
Calling any other functions in libc never change the internal state of
the mblen(), except for calling setlocale(3) with the LC_CTYPE category
changed to that of the current locale. Such setlocale(3) calls cause the
internal state of this function to be indeterminate.
Regards.
--
Sebastien Marie
