pghmcfc
commented
2 years ago This seems to be related to the behaviour of Crypt::CBC when encrypting an empty block. If I make this change to Crypt::CBC:
diff -up ./lib/Crypt/CBC.pm.pws3 ./lib/Crypt/CBC.pm
--- ./lib/Crypt/CBC.pm.pws3 2022-02-16 14:47:01.022412974 +0000
+++ ./lib/Crypt/CBC.pm 2022-02-16 14:52:27.617799760 +0000
@@ -252,7 +252,7 @@ sub finish (\$) {
$result = $self->{padding}->($result,$bs,'d') if $self->_needs_padding;
} else {
$block = $self->{padding}->($block,$bs,'e') if $self->_needs_padding;
- $self->$code($self->{crypt},\$iv,\$result,[$block]) unless length $block==0 && !$self->_needs_padding
+ $self->$code($self->{crypt},\$iv,\$result,[$block]) unless length $block==0;
}
delete $self->{'civ'};Then the Crypt-PWSafe3 test runs OK (assuming the -blocksize options are removed as discussed earlier). However, Crypt-CBC's own test suite then fails:
$ make test
PERL_DL_NONLAZY=1 "/usr/bin/perl" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_harness(0, 'blib/lib', 'blib/arch')" t/*.t
t/AES.t .............. ok
t/Blowfish.t ......... ok
t/Blowfish_PP.t ...... ok
t/CAST5.t ............ skipped: Crypt::CAST5 not installed
t/DES.t .............. ok
input must be 8 bytes long at /usr/lib64/perl5/vendor_perl/Crypt/Blowfish.pm line 58.
t/func.t .............
Dubious, test returned 255 (wstat 65280, 0xff00)
Failed 823/1537 subtests
t/IDEA.t ............. ok
using Crypt::Cipher::AES for testing
t/nopadding.t ........ ok
t/null_data.t ........ ok
t/OFB.t .............. ok
using Crypt::Cipher::AES for testing
t/onezeropadding.t ... ok
t/parameters.t ....... ok
t/pbkdf.t ............ ok
t/PCBC.t ............. ok
Using Crypt::Cipher::AES for test
t/preexisting.t ...... ok
t/Rijndael.t ......... ok
t/Rijndael_compat.t .. ok
Test Summary Report
-------------------
t/func.t (Wstat: 65280 Tests: 716 Failed: 2)
Failed tests: 205, 461
Non-zero exit status: 255
Parse errors: Bad plan. You planned 1537 tests but ran 716.
Files=17, Tests=4238, 15 wallclock secs ( 0.10 usr 0.01 sys + 14.66 cusr 0.22 csys = 14.99 CPU)
Result: FAIL
Failed 1/17 test programs. 2/4238 subtests failed.
At this point I'm not sure whether the problem is in Crypt-CBC or Crypt-PWSafe3. It seems reasonable to me that a padded empty block (which in the case of Crypt-PWSafe3 will just be an empty block anyway since a null-padded empty block is an empty block) should be encrypted, which is what Crypt-CBC currently does.
TLINDEN
lstein
Hi,
I am the author of Crypt::PWSafe3, a module to read and write passwordsafe3 files. Until 2.33 of
Crypt::CBCeverything worked as expected, but starting with 3.01 (and up) it doesn't anymore. There's an issue in my module (https://github.com/TLINDEN/Crypt--PWSafe3/issues/12) and another on RT for yours (https://rt.cpan.org/Public/Bug/Display.html?id=134355).The problem is, if I remove all references to
-blocksize, then it should work, since Crypt::Twofish propagates a blocksize of 16, which I had in my module as well. So, removing this call should fix it. But it doesn't. pwsafe3 files created with Crypt::CBC >= 3.01 are not decryptable with Crypt::CBC >= 3.01 anymore. However, files created with Crypt::CBC < 3.01 are still readable and stay intact as long one doesn't edit them (which re-creates the file, now written with 3.01, which can't be decrypted).I don't know if anything else has changed. But I looked into the git history of the module and discovered that you made a refactoring of the code. Maybe somehing went wrong during the process.
If you want to reproduce the problem, checkout https://github.com/TLINDEN/Crypt--PWSafe3, comment out all 3 references to
-blocksizeinlib/Crypt/PWSafe3.pmand runperl Makefile.PL && make test. You'll realize that test 2 succeeds, it opens the supplied 9 year old pwsafe3 file successfully. Everything breaks apart from test 3 and following, they all create new pwsafe3 files.