crypto: throw in setAuthTag on invalid length

[tniessen]

The current implementation performs limited checks only and silently ignores superfluous bytes of the authentication tag. This change makes setAuthTag throw when

• the user-specified authTagLength does not match the actual tag length, especially when the authentication tag is longer than 16 bytes, and when
• the mode is GCM, no authTagLength option has been specified and the tag length is not a valid GCM tag length.

This change makes the conditional assignment in SetAuthTag unnecessary,
which is replaced with a CHECK.

Refs: #17825

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• commit message follows commit guidelines

[tniessen]

CI: https://ci.nodejs.org/job/node-test-pull-request/14300/

[mscdex]

/cc @nodejs/crypto

[yhwang]

LGTM

[yhwang]

but I guess we should throw here, which would make this a breaking change.

+1 for throwing an exception.

and yes, old implementation truncate the authtag......

[BridgeAR]

Is this ready to land?

[tniessen]

@BridgeAR No, I missed an edge case which @yhwang discovered, so this could cause a regression. The only solution seems to be to throw an error if the input value is too long, I will update the PR as soon as I have the time to do so.

[richardlau]

As per #20040 (comment).

[tniessen]

This patch turned out completely different than I had originally planned, but it still goes into the correct direction in my opinion. Please take another look.

As this change is now semver-major anyway, I also unified related error messages.

^{I also took the liberty of marking the discussion of the old change as "Outdated", I can't find the issue about when to use that feature. Feel free to unhide the discussion if this was incorrect.}

[bnoordhuis]

LGTM at a quick glance.

[tniessen]

New CI: https://ci.nodejs.org/job/node-test-pull-request/15168/

I know that most of the TSC is probably travelling right now, but still pinging @nodejs/tsc as this is a semver-major change now. Also @richardlau due to his blocking review.

outdated

[tniessen]

Only failure seems to be infra-related:

/var/folders/dr/m8hrvy3d71172187rtx7vm380000gp/T/jenkins12712608435428624123.sh: line 6: /Users/iojs/Library/Python/2.7/bin/tap2junit: No such file or directory

Here's CitGM: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/1446/

[Trott]

Re-running node-test-commit-osx: https://ci.nodejs.org/job/node-test-commit-osx/18957/

[tniessen]

CitGM looks good as far as I can tell. Landed in faf449c.