On Tue, 02.09.08 09:31, rdiezmail-pulseaudio at yahoo.de (rdiezmail-pulseaudio at yahoo.de) wrote: heya! > I have noticed that the assert() calls inside the libpulse client > library are enabled by default. I would expect asserts to be > disabled in release builds, via the NDEBUG macro or similar > compile-time switch. > > What's the common practice for asserts in the OSS world in this kind > of libraries? Disabling asserts in production builds is a bad idea. If you like to shoot yourself in the foot, you're welcome to do so, but I see no reason to ship support for this upstream. Passing CLFAGS=-DNDEBUG should be enough. Defining NDEBUG is something for Gentoo-kiddies, something that would fit well into Jorge's awesome http://funroll-loops.info web site. ;-) The perfomance impact of the asserts is absolutely minimal, they don't appear in inner loops. Especially in software that is accessible from the network or to other users (which is the case for PA) it's better to be safe than sorry -- checking once too often whether everything is running correctly is better than once too seldom. Software is always buggy. Catching errors early and with somewhat explanatory error messages is of big benefit in contrast to just segfaulting or losing data. And this truth doesn't change if you are using debug builds or production builds. Most distros leave asserts in the production code (except maybe Gentoo, but they smoke crack anyway ;-)). And it is good that way. Lennart -- Lennart Poettering Red Hat, Inc. lennart [at] poettering [dot] net ICQ# 11060553 http://0pointer.net/lennart/ GnuPG 0x1A015CC4
