[Development] qDebug, qWarning, qCritical, qFatal in Qt source code

[Giuseppe D'Angelo]

On 3 May 2014 12:28, Kurt Pattyn <pattyn.kurt at gmail.com> wrote:
> Recently I had to remove qWarning statements from a submit (for very plausible reasons), but a quick search through qtbase revealed a lot of qWarning statements.

> So are there any rules of thumb?

Please also cf. the other thread about using categorized logging from within Qt.

It's never good to use qDebug unconditionally, as that's useless
information for the users. qWarning, qCritical, qFatal are fine to
indicate a major problem, usually originating from user code or from
some specific configuration issue.
For instance a failing QObject::connect will print a warning through
qWarning. Running a setuid binary will trigger a qFatal. Q_ASSERTs go
through qFatal if they fail (but they're compiled out in release
mode).
qCritical is not really used in Qt.

qFatal on its own will crash the program; qWarning / qCritical can be
set to do the same by exporting QT_FATAL_WARNINGS=1. In programs we
work on we strive to get 0 warnings by keeping that env variable set
at all times.


About qDebug: usually in Qt source code we had debug statements
wrapped in macros:

[snip]
#ifdef ENGINE_DEBUG
qDebug() << "Foo" << foo;
#endif
[snip]

or:

[snip]
#ifdef ENGINE_DEBUG
#define engineDebug qDebug
#else
#define engineDebug QT_NO_QDEBUG_MACRO
#endif

engineDebug() << "Foo" << foo;
[snip]

But for new code we should use categorized logging and its tests.

Hope this helps,
-- 
Giuseppe D'Angelo