On Wed, 15 Apr 2015 11:07:16 +0200 Nicolas George <geo...@nsup.org> wrote:
> Le sextidi 26 germinal, an CCXXIII, wm4 a écrit : > > Because we want to test the API, not the command line tools. > > And the command line tools, as everybody knows, operate by magic instead of > using the API. The command line tools are only 1 API user out of hundreds. > If there is a point of the API that can not be accessed with the > command-line tools, fix that first. Not possible. The command line tools behave in a very specific way. It can't test the API, it can only test the command line tools' way they use the APIs. They don't necessarily use the full API, and they can mask problems in the API. I'm not sure how your suggestion can even taken seriously, since the tools tend to mess with API internals, use some internal headers, and the libraries frequently tend to contain hacks to make up for deficits in the tools. (MPlayer is the only other API user which gets this special treatment.) There's just no point in testing a single API user. We need these API tests, if only for the fact that these "artificial" tests may be able to cover 100% of the API, something no real API user would do. Your attitude also shows an amazing chauvinism: who cares about the API users (and any downstream projects), as long as the command line tools are fine? Because your world view seems to be restricted to the ffmpeg git repo. _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel