D24980: Port ssl_cert_errors meta data from KSslError to QSslError

[Volker Krause]

vkrause added inline comments.

INLINE COMMENTS

> ahmadsamir wrote in jobuidelegate.cpp:365
> Nitpick, wouldn't QLatin1String be better here?
> 
> c.f. Marc Mutz's talk about QStringLiteral and QLatin1String: https://youtu.be/Ov7s0GgBbOQ?t=2806

sslMetaData is a QMap<QString, ...>, ie. value() has no QLatin1String overload, calling it with a QLatin1String will work but convert to a QString at runtime (involving a memory allocation), using QStringLiteral avoids that.

> ahmadsamir wrote in ksslinfodialog.cpp:263
> Nitpick, shouldn't there be a "// static" comment like the other method?
> 
> Wouldn't a QVector be better here (QVector<QList<QSslError::SslError>>)? that is IIUC what the QList docs and [1] are saying.
> 
> [1]https://marcmutz.wordpress.com/effective-qt/containers/

In theory, yes. However, changing this here and now would trigger quite some ripple effect through the code, which currently is using QList everywhere, so I'd suggest to do that as a separate change (if at all), to not hide security-relevant changes under lots of QList -> QVector porting noise. The other thing to consider the consistency in this code if we change just a single use from QList to QVector, especially if that's not a performance-sensitive path and Qt6 is supposed to change all this anyway (QList's implementation going away).

REPOSITORY
  R241 KIO

BRANCH
  next

REVISION DETAIL
  https://phabricator.kde.org/D24980

To: vkrause, nicolasfella
Cc: ahmadsamir, nicolasfella, kde-frameworks-devel, LeGast00n, GB_2, michaelh, ngraham, bruns
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-frameworks-devel/attachments/20191109/39ef466c/attachment-0001.html>