Bug 173253 – [GCrypt] Use utility functions in CryptoKeyEC, CryptoKeyRSA

• Component: New Bugs
• Status: RESOLVED FIXED

[Zan Dobersek]

[GCrypt] Use utility functions in CryptoKeyEC, CryptoKeyRSA

[Zan Dobersek]

Created attachment 312647 [details]
Patch

[Michael Catanzaro]

Comment on attachment 312647 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=312647&action=review

Nice!

> Source/WebCore/crypto/gcrypt/CryptoKeyECGCrypt.cpp:227
> +    return WTFMove(q.value());

Why do you use WTFMove() here? Doesn't this prevent return value optimization?

> Source/WebCore/crypto/gcrypt/CryptoKeyRSAGCrypt.cpp:70
> +    return WTFMove(data.value());

Ditto.

[Zan Dobersek]

Comment on attachment 312647 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=312647&action=review

>> Source/WebCore/crypto/gcrypt/CryptoKeyECGCrypt.cpp:227
>> +    return WTFMove(q.value());
> 
> Why do you use WTFMove() here? Doesn't this prevent return value optimization?

RVO doesn't help here because q is a std::optional<> object. `q.value()` exposes the embedded Vector<> object, but returning just that would result in copying the object. So the embedded value has to be casted to an rvalue reference via WTFMove() to enforce the move constructor to be generated.

So the Vector<> object inside std::optional<> is moved into the return value, and after that std::optional<> deletes the emptied-out Vector<> it still holds.

[Zan Dobersek]

Comment on attachment 312647 [details]
Patch

Clearing flags on attachment: 312647

Committed r218100: <http://trac.webkit.org/changeset/218100>

[Zan Dobersek]

All reviewed patches have been landed.  Closing bug.