Re: [Libexif-devel] Error handling

[Joerg H. <jo...@de...>]

On Mon, May 17, 2004 at 08:18:56PM +0200, Lutz Müller wrote:
> On Sun, 2004-05-16 at 16:17, Joerg Hoh wrote:
> > I prefer a to replace at least some void-functions with (signed) int with a
> > standard return value of 1; if an error happens, they will return something
> > != 1 (normally < 1).
> 
> (1) User clicks and empty information shows up.
>     This is the case now.
> 
> (2) User clicks and gets message: "An error occurred."
>     This is your proposal.

Nope. I don't talk about any kind of interaction with the user on
application side. I talk about how errors in functions can be detected from
a calling one.

> (3) User clicks and gets message: "Error xy occurred."
>     Like EXIF_ERROR_CORRUPT_DATA, 
>          EXIF_ERROR_UNKNOWN_FORMAT
> 
> (4) User clicks and gets message: 
>     "Error xy occurred at byte 4711 in EXIF data. 
>      The EXIF data in your image is broken. 
>      Please contact mailing list xyz."
>     Implemented i.e. like
>     void exif_data_load_data (..., ExifResult *result);
>     where 
>     typedef struct {
>         char *msg;
>         ExifError err;
>     } ExifResult;

I don't want to bloat the library. And yes, such a solution would be bloat.
Why do I have to return a struct when no error occurred? How do I allocate
a structure when a malloc(10) failed? Then we should have an error handling
for the error handling...

Returning a simple int in some functions is just enough. On high-level
functions (which will be be called from application side) we can think
about such a solution, but not on low-level.

Joerg

-- 
Fachbegriffe der Informatik (Nr 275): Luster-Format
- Positiv gemeint: Eßfreudig Negativ gemeint: Das Äquivalent von zwei Öltanks

	Alexander Stielau