No -Werror in CFLAGS
No -Werror in CFLAGS
Posted Feb 20, 2010 1:30 UTC (Sat) by giraffedata (guest, #1954)Parent article: FOSDEM'10: distributions and downstream-upstream collaboration
There should also be no -Werror in CFLAGS, as the project should respect the user's choice
I don't see where -Werror is a user's choice issue.
I learned early in my GNU-using career not to put -Werror in distributed code because it causes lots of needlessly failing builds because of insignificant differences between the user's and the distributor's systems (like compiler version) and causes virtually no productively failing builds.
I do use -Werror unfailingly on my own builds of things that I distribute, though, and make sure there are no warnings on my own system. Maybe that's the user's choice the statement means. But I see no point whatsoever to choosing to add -Werror to a build of someone else's code -- that's just daring the build to fail.
Posted Feb 21, 2010 13:43 UTC (Sun)
by Darkmere (subscriber, #53695)
[Link]
There has been plenty of times where I've been forced to mangle Makefiles because the developer had -Wall -Werror in their Makefiles, and built with an older/different/odder version of GCC/compiler-of-choice than what we did, and thereby causing a ton of failures.
Compilers become pickier. When you release a piece of software to the wild, -Wall and -Werror do not belong in the sources, as two months later, your tarball will fail to build against the current cvs version of GCC, because it suddenly warns in -Wall that you don't have your commas aligned vertically with your paranthesis, so the smileys turn the wrong way.
The point is, be strict when debugging and developing, be lenient to downstream, don't force downstream to patch Makefiles, configure files and similar just because you ship with -Wall -Werror -D_FAIL_ON_GLIBC22
No -Werror in CFLAGS
( Or similar . )
