Re: [Libexif-devel] Error handling

[Lutz <lu...@us...>]

On Tue, 2004-05-18 at 13:51, Joerg Hoh wrote:
> I think at this moment it would be easier to
> work with return values. The errors I want to catch with this approach are
> relativly "simple" errors like error return values of libc-calls (malloc,
> realloc) and possibly invalid (NULL) arguments. Look at the points in
> functions where at this moment we just do a simple "return". Most times
> it's enough to distinct between "error occured" and "no error".

I had an hour to play with libexif and came up with a proposal for
debugging messages (exif-log.[c,h], try exif --debug
non_existing_file.jpg).

API didn't change. But now you have the possibility to watch what is
going on internally using a callback-function.

While thinking about how that could be used to satisfy your needs, I ran
out of ideas regarding the following question:

At which point does an invalid argument or a failed malloc become fatal?
At libexif level? At the maker note level? I don't think a bug in the
maker note code should let the loading of the ExifData fail. There seems
to never be a 1/0-decision.

Regards

Lutz