Invalid MAC with i386

[pmaziere]

For specific reasons , I have to work with an i386 computer running debian testing without iibsodium installed.
While giving a try to Paseto, I was hit with an "Invalid MAC" error when parsing the token, which lead me to #38.

After running the phpunit tests on current master HEAD, here is the final output:

Time: 1.8 hours, Memory: 18.00MB

There was 1 error:

1) FileTest::testSeal
SodiumException: Invalid MAC

/srv/www/local/sodium_compat/src/Crypto32.php:1003
/srv/www/local/sodium_compat/src/Crypto32.php:664
/srv/www/local/sodium_compat/src/Crypto32.php:519
/srv/www/local/sodium_compat/src/Compat.php:983
/srv/www/local/sodium_compat/tests/unit/FileTest.php:147

--

There was 1 failure:

1) FileTest::testBox
Failed asserting that two strings are identical.
--- Expected
+++ Actual
@@ @@
-d1be33c603b2363bc2cda62795b458d34aba8d03a9c190fa5fe0c14603f57eb6b0db46a322a21fdef20498304d2fbc9d
+46679e2378061e778f9b1cf4f796f59822cf39d3017fc62464d5be2e2dbabd5648593477ef571af3732ff33dacf116af

/srv/www/local/sodium_compat/tests/unit/FileTest.php:43

ERRORS!
Tests: 173, Assertions: 6699, Errors: 1, Failures: 1, Skipped: 58.

I know that i386 targets are not high priority, but would it be possible for someone to have a look at these tests. Hopefully, it may result in a quick and easy fix.

thanks

[paragonie-scott]

I'm worried this might be a Curve25519 bug, which is the slowest thing to test on 32-bit.

[paragonie-scott]

Can you let me know if https://github.com/paragonie/sodium_compat/releases/tag/v1.6.4 fixed the issue for you?

[pmaziere]

Still having the invalid MAC error in FileTest::testSeal, but the FileTest::testBox is now passed successfully

1) FileTest::testSeal
SodiumException: Invalid MAC

/srv/www/local/sodium_compat/src/Crypto32.php:1003
/srv/www/local/sodium_compat/src/Crypto32.php:664
/srv/www/local/sodium_compat/src/Crypto32.php:519
/srv/www/local/sodium_compat/src/Compat.php:983
/srv/www/local/sodium_compat/tests/unit/FileTest.php:147

ERRORS!
Tests: 173, Assertions: 6704, Errors: 1, Skipped: 58.

[pmaziere]

After installing a VM on a 64 bits intel computer, I confirm that tests are passed successfully for this architecture. But I have the same 'Invalid MAC' error as described in my initial post with my Paseto attempt, therefore in addition to this bug report, I guess I must have something wrong in my paseto related code.
And I can't affirm that my Paseto 'invalid MAC' error is related to the sodium_compat FileTest::testSeal 'Invalid MAC' error.

[paragonie-scott]

It wouldn't be. PASETO doesn't touch the File API at all.

[pmaziere]

Ok, I confirm that the PASETO issue was unrelated to any sodium_compat bug. It is now fixed.

Anyway, I keep this bug opened since the FileTest::testSeal error is still pending. Maybe someone else could try to reproduce the bug, so that an isolated device specific issue can be ruled out.

[paragonie-scott]

This fixes the unit tests on my Windows machine: c9ef28e

[paragonie-scott]

This is fixed in v1.6.5.

[pmaziere]

Not here (debian testing 64bit hosting an LXC guest running debian stretch 32bit)

1) FileTest::testSeal
SodiumException: Invalid MAC

/srv/www/local/sodium_compat/src/Crypto32.php:1003
/srv/www/local/sodium_compat/src/Crypto32.php:664
/srv/www/local/sodium_compat/src/Crypto32.php:519
/srv/www/local/sodium_compat/src/Compat.php:983
/srv/www/local/sodium_compat/tests/unit/FileTest.php:147

[paragonie-scott]

Okay, I'm not able to reproduce this in v1.6.5 on my setup. I'll look into setting up a congruent system so I can study it further.

[pmaziere]

Since you were not able to reproduce it, I cloned the repository from scratch and ran the tests again, with a fully successfull output this time.
Thanks for the time you spent on this bug, and for all your work.