Re: critical warnings; turn them off now?

On 3/8/06, Behdad Esfahbod <behdad cs toronto edu> wrote:
> On Tue, 7 Mar 2006, Federico Mena Quintero wrote:
> > A critical warning is not some annoying spew in your console which you
> > can ignore; it is an indication that something has gone HORRIBLY wrong
> > and you should fix it immediately.  It is not a faint light that says
> > "check engine", it is a big siren screaming, "HOLY SHIT, SOMEONE PUT
> In many cases it's a false alarm.  Something that should not have
> marked as such in the first place.  From my experience with Pango
> modules recently, one major cause of these false alarms are
> g_return_if_fail()s.  It's important to differentiate between
> some unusual but perfectly valid failure of something (like
> failing to lock a font face, because the font file may have been
> removed), and invalid input from the user.  With former, you may
> want to do a g_warning and return, only with latter one should
> use g_return_if_fail.
> In other words, you should use g_return_if_fail in cases that
> upon seeing the crash, the developer has something to fix.  This
> is not quite the case when font locking fails :)

And the big siren wailing at you will give a good push to fix it to
behave correctly in an erroneous situation (it is a serious bug to
handle error cases badly, even if it in this case is maybe just a
wrong message type).

At least that was the idea if I understood it right. :)

False alarms should be the easiest ones to fix, so they have extra
reason to be pushed for fix in my opinion...

Bill Haneman said:
> The a11y critical warnings are 99% false alarms, due to misuse of
> g_return_*.  Unfortunately, fixing this is very time-consuming since
> each instance of g_return_* in the a11y code must be examined
> ndependently in order to determine whether it's a "false alarm" or not.

And you can probably guess how many of those would be fixed if there
wasn't a gun pointed at the developers head with an ultimatum "fix it
or crash" :)

It's pain in the arse, but if it makes code better and more correct,
it should it be worth it, right?

Kalle Vahlman, zuh iki fi
Powered by
Interesting stuff at

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]