See http://cgit.freedesktop.org/xorg/util/macros/commit/xorg/macros.m4.in?id=b8a13e531db208a7c4c9a43b74b196104e7c66ed Multiple users on NetBSD have seen like: checking if cc -std=gnu99 supports -Wredundant-decls... yes checking if cc -std=gnu99 supports -Werror=implicit... no checking if cc -std=gnu99 supports -errwarn=E_NO_EXPLICIT_TYPE_GIVEN -errwarn=E_NO_IMPLICIT_DECL_ALLOWED... eval: 1: Syntax error: Bad substitution *** Error code 2 For the generated configure, a workaround patch is quoting it: - cacheid=`$as_echo xorg_cv_cc_flag_-errwarn=E_NO_EXPLICIT_TYPE_GIVEN -errwarn=E_NO_IMPLICIT_DECL_ALLOWED` + cacheid=`$as_echo "xorg_cv_cc_flag_-errwarn=E_NO_EXPLICIT_TYPE_GIVEN -errwarn=E_NO_IMPLICIT_DECL_ALLOWED"` Some explanation is at http://mail-index.netbsd.org/tech-pkg/2012/06/05/msg009446.html The macro for configure.ac has: XORG_TESTSET_CFLAG([[BASE_]PREFIX[FLAGS]], [-Werror=implicit], [-errwarn=E_NO_EX PLICIT_TYPE_GIVEN -errwarn=E_NO_IMPLICIT_DECL_ALLOWED]) Regardless if this is a bug or not in the sh (note this sh is used on non-NetBSD platforms too), is it okay to split up the check to two macros? diff --git a/xorg-macros.m4.in b/xorg-macros.m4.in index d589ebd..d122edf 100644 --- a/xorg-macros.m4.in +++ b/xorg-macros.m4.in @@ -1646,7 +1646,8 @@ XORG_TESTSET_CFLAG([[BASE_]PREFIX[FLAGS]], [-Wredundant-decls]) # when there are problems that should be fixed. if test "x$SELECTIVE_WERROR" = "xyes" ; then -XORG_TESTSET_CFLAG([[BASE_]PREFIX[FLAGS]], [-Werror=implicit], [-errwarn=E_NO_EXPLICIT_TYPE_GIVEN -errwarn=E_NO_IMPLICIT_DECL_ALLOWED]) +XORG_TESTSET_CFLAG([[BASE_]PREFIX[FLAGS]], [-Werror=implicit], [-errwarn=E_NO_EXPLICIT_TYPE_GIVEN]) +XORG_TESTSET_CFLAG([[BASE_]PREFIX[FLAGS]], [-Werror=implicit], [-errwarn=E_NO_IMPLICIT_DECL_ALLOWED]) XORG_TESTSET_CFLAG([[BASE_]PREFIX[FLAGS]], [-Werror=nonnull]) XORG_TESTSET_CFLAG([[BASE_]PREFIX[FLAGS]], [-Werror=init-self]) XORG_TESTSET_CFLAG([[BASE_]PREFIX[FLAGS]], [-Werror=main])
Just to clarify: Shell variables aren't allowed to have spaces in their names.
I'd rather XORG_TESTSET_CFLAG use an appropriate environment variable as the solution. The suggested change would result in -Werror=implicit appearing twice.
Adding Jon, since this is fallout from the result caching changes.
OBATA Akio writes that the problem may have been solved in the meantime: http://mail-index.netbsd.org/pkgsrc-users/2012/06/09/msg016498.html
If I'm understanding correctly, the current version of xorg-macros doesn't trip over the shell bug which caused this to fail on NetBSD. So, I don't think any further action needs to be taken and I'm marking this bug as resolved. Please re-open this bug if my understanding is defective. (In reply to comment #0) > Regardless if this is a bug or not in the sh (note this sh is used on > non-NetBSD platforms too), is it okay to split up the check to two macros? I can't really speak to if this is desirable. At the moment, we do one check to see if the compiler supports both flags.
> (In reply to comment #0) > > Regardless if this is a bug or not in the sh (note this sh is used on > > non-NetBSD platforms too), is it okay to split up the check to two macros? > > I can't really speak to if this is desirable. At the moment, we do one check > to see if the compiler supports both flags. To answer the question, one must follow the evolution of the macros as changes need to be backward compatible. Older xorg packages must behave the same way as they did when built with the latest xorg-utils package. Another theoretical reason is for functional decomposition to improve reusability.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.