On date Monday 2015-03-16 11:11:55 +0100, Clément Bœsch encoded: > On Mon, Mar 16, 2015 at 11:05:12AM +0100, Stefano Sabatini wrote: > > On date Monday 2015-03-16 11:00:12 +0100, Stefano Sabatini encoded: > > > On date Sunday 2015-03-15 15:15:34 +0100, Clément Bœsch encoded: > > [...] > > > > To elaborate on this, the bug here is referring to an allocation check > > > > not > > > > done in the caller (there are many currently since I'm just introducing > > > > the error handling). > > > > > > I won't block this patch, but getting a crash in a specified point of > > > the program is more useful than failing with this: > > > > > > > > A bug occurred somewhere > > > > Now this is not correct, it should be: > > there is a bug somewhere > > > > (since bug don't "occurr" but stay/are). Now that (the presence of a > > "bug") is not even a good reason to fail a program, it should had been > > named something more meaningful like "unexpected condition" (but again > > - for that a crash or an assert is just more useful to the > > programmer). > > We are talking about a crash non programmer would get. If you want, I can > add an av_log message saying that the filter is buggy. I just don't want > to create dozens of potential crashes when we know this can be NULL in the > current state. Asserting on the result of malloc feels just wrong.
What about asserting in the callee if the argument is NULL? -- FFmpeg = Furious and Freak Monstrous Pure Erroneous God _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
