[gmime-devel] Issue with header decoding

[evil legacy <evil legacy gmail com>]

Hey all,

I came across this header: "=?windows-1255?Q?=E1?=" which should decode to a single character, 

in the following reproduction code, decoding s0 returns an empty string and only after adding "=00" i get the expected result.

<code>

    char *s0 = "=?windows-1255?Q?=E1?=";

    char *s1 = "=?windows-1255?Q?=E1=00?=";

    printf("'%s' -> '%s'\n", s0, g_mime_utils_header_decode_text(s0));

    printf("'%s' -> '%s'\n", s1, g_mime_utils_header_decode_text(s1));

</code>

# ./gmime_decode 

'=?windows-1255?Q?=E1?=' -> ''

'=?windows-1255?Q?=E1=00?=' -> 'ב'

I've tried both gmime-2.4 and 2.6, on 3 different linux platforms (gentoo,arch,centos) and after digging in the sources, I noticed that iconv behaves unexpectedly when asked to decode a single char. follows code to confirm:

<code>

#include <stdio.h>

#include <iconv.h>

#include <stdlib.h>

int main(int argc, char **argv)

{

    iconv_t cd = iconv_open("utf-8","windows-1255");

    

    int insize = strtol(argv[1],0,10);

    char inbuf[2] = { 225, 0 };

    char outbuf[10];

    char *inptr = inbuf;

    char *wrptr = outbuf;

    size_t avail = 10;

    int rc = iconv (cd, &inptr, &insize, &wrptr, &avail);

    printf("rc: %d, avail: %d, decoded: %s\n", rc, avail, outbuf);

    return 0;

}

</code>

result:

# gcc test.c 

# ./a.out 1

rc: 0, avail: 10, decoded: ü§¿%Þá

# ./a.out 2

rc: 0, avail: 7, decoded: ב

I've noticed that in gmime-utils.c such behavior occurs when decoding the header in question:

i.e, declen=1

<code gmime-utils.c>

size_t n = charset_convert (cd, (char *) decoded, declen, &buf, &len, &ninval);

</code>

What's the recommended tactic resolving this issue?

Should I detect and fix such headers before trying to decode them?

Best Regards.