perl-unicode

Re: UTF8 behavior under -T (Taint) mode

2004-01-01 07:30:05
On Jan 01, 2004, at 21:49, Masanori HATA wrote:
Sorry, no. Since the case which I would like to suggest
seems not to be fatal. Perl would not die, but it would
take the tainted value as a Non-UTF8 string.

My sample code is like below (test.pl):
-------------------------------------------------
utf8::decode(my $text0 = "\x{3042}"  ); # clean
utf8::decode(my $arg   = $ARGV[0]    ); # tainted
utf8::decode(my $text1 = "$arg$text0"); # tainted
utf8::decode(my $text2 = "$text0$arg"); # tainted

print length($text1), "\n";
print length($text2), "\n";
-------------------------------------------------

Aha!  I see your point at last.  And I found your argument was correct.

When I run this code with 'perl -T test.pl a', the result is:

To clear your point, I have modified your script with Devel::Peek. Pay attention to the $text1 result.

without -T
% perl test.pl a
SV = PV(0x812354) at 0x80a960
  REFCNT = 1
  FLAGS = (PADBUSY,PADMY,POK,pPOK,UTF8)
  PV = 0x428090 "a\343\201\202"\0 [UTF8 "a\x{3042}"]
  CUR = 4
  LEN = 5
2
SV = PV(0x812e10) at 0x80f2a8
  REFCNT = 1
  FLAGS = (PADBUSY,PADMY,POK,pPOK,UTF8)
  PV = 0x405150 "\343\201\202a"\0 [UTF8 "\x{3042}a"]
  CUR = 4
  LEN = 5
2

with -T
% perl -T test.pl a
SV = PVMG(0x819a88) at 0x80a954
  REFCNT = 1
  FLAGS = (PADBUSY,PADMY,GMG,SMG,pPOK)
  IV = 0
  NV = 0
  PV = 0x428540 "a\343\201\202"\0
  CUR = 4
  LEN = 5
  MAGIC = 0x405480
    MG_VIRTUAL = &PL_vtbl_taint
    MG_TYPE = PERL_MAGIC_taint(t)
    MG_LEN = 1
4
SV = PVMG(0x819af4) at 0x80f69c
  REFCNT = 1
  FLAGS = (PADBUSY,PADMY,GMG,SMG,pPOK,UTF8)
  IV = 0
  NV = 0
  PV = 0x4054e0 "\343\201\202a"\0 [UTF8 "\x{3042}a"]
  CUR = 4
  LEN = 5
  MAGIC = 0x4010d0
    MG_VIRTUAL = &PL_vtbl_taint
    MG_TYPE = PERL_MAGIC_taint(t)
    MG_LEN = 1
2

I am not sure how severe it is but this is a bug indeed.

(My system is perl5.8.1 MSWin32-X86-multi-thread)

I have duplicated the result with Perl 5.8.2 on Mac OS X as well as maintperl(_at_)21987 on FreeBSD. And using Encode::decode_utf8 does not help either because it simply calls utf8::decode. And you can't use Encode::decode("utf8", ...) in this particular case because Encode::decode() checks and clobbers at "Cannot decode string with wide characters". Hmm....

Dan the Perl5 Porter

<Prev in Thread] Current Thread [Next in Thread>