On Tue, 2004-12-14 at 17:52 +0000, Gustavo J. A. M. Carneiro wrote: > I don't see any advantage in hardcoding -Werror in makefiles, as > opposed to let developers add it to their own CFLAGS. I see two advantages. First, it makes it less likely for patch contributors to send buggy patches. For example Rhythmbox has defaulted to -Werror with a bunch of other flags such as -Wdeclaration-after-statement and -Wmissing-prototypes for a long time. Some people joined IRC to ask about this because they hadn't encountered warnings like -Wdeclaration-after-statement before. It's much faster for them to fix it during development then it would be for the maintainers to reject the patch and have the contributor redo it, or for the maintainer to spend time fixing these things. Second, while one can get spurious errors, often the errors caught by -Werror are ones that might not show up on a particular developer's machine; e.g. 64-bit pointer/int bugs. Breaking the compilation makes it much more likely that a user compiling from CVS/arch/whatever who happens to have a 64-bit machine will see the bug instead of just having a random warning scroll by.
Attachment:
signature.asc
Description: This is a digitally signed message part