Re: Problem with localized

[David Chisnall]

On 29 May 2011, at 14:07, Dr. H. Nikolaus Schaller wrote:

> just guessing: %s is the formatting specified for a C string while %@ expects 
> an NSString...
> 
> If %s is applied to a NSString, it tries to interpret the iVars of the 
> NSString object as characters, aka "garbage".


If you're lucky, and there is a 0 byte somewhere near the start of the object - 
otherwise it will crash.

When using a recent compiler, GNUstep sets some attributes that make it 
possible to check this sort of thing at compile time.  The correct way of doing 
this is to use the English (or your favourite other language) string as the 
name of the string, so you do:

        [NSString stringWithFormat: _(@"%s seems to be running for the first 
time.\nPlease check the settings before you commence work"), APP_NAME],

The method that _() expands to declares that any string that it may be used as 
a format string anywhere that the input string may be used.  You'd then get a 
compile-time warning saying that %s expects a C string but you provided it with 
an object.  Similarly, you'd have see an obvious error if your strings file 
contained something like this:

"%@ some English words" = "%s some other words"

You should actually be seeing an error with this usage, saying that the second 
argument to  -stringWithFormat: is an object but there is no corresponding 
format specifier.  I'm a bit surprised that you don't.  The usage 
@"AppController.running" is wrong, since this lookup function will return its 
input if it can't find the strings file.

We could probably add some extra checking in the NSBundle code that looks up 
the format strings to ensure that the input and output format strings match.  
Alternatively, we could add a verify_strings tool that -make could run 
automatically, checking format string agreement.

David

-- Sent from my PDP-11