Description
I like the idea of your program, but I do have a few suggestions and ideas. IMO it should be a little bit more Unix like. Simple invocation, scriptable, ...
Currently you get output that has to be redirected to /dev/null if you want to use it in a script. Nobody really cares about text output in such a case. But for this program there's no text required at all. (Maybe you wanted to play with multiple languages, but it's just not necessary and superfluous output only annoys people - or at least it annoys me.)
Arguments are great, but there should be short and long options. e.g. -f for --filename, ...
The only important argument is the size, thus it should not need an argument at all. You shoud be able to create a file by just doing this:
$ trash-dump 1048576
and you get a 1 megabyte file called tdump.tdp (or whatever you want to choose as the default filename).
Which brings me to my next suggesttion. Allow modifiers like k, m, g, which multiply the size by a factor.
10m would be 10 * 1048576 bytes = 10 megabytes.
Also, the program should not output any text, unless invoked with -v or --verbose. Text output can be nice for statistics or other stuff, but there's nothing needed in the program. You can use time (Unix/Linux command) to measure the runtime. The size of the file was specified at input time. So why do we need a text output?
The return code usually tells me everything I want to know. 0 for success. 1 for error. You can use different error codes for different errors. (disk full = 2, file exists = 1, unspecified = 3, ....).
P.S.: A Makefile would be great to build and install the program.