]> git.ipfire.org Git - thirdparty/gcc.git/commit - gcc/c-family/ChangeLog
Move #undef DEF_BUILTIN* to builtins.def
authorRichard Sandiford <richard.sandiford@arm.com>
Sat, 7 Nov 2015 10:18:38 +0000 (10:18 +0000)
committerRichard Sandiford <rsandifo@gcc.gnu.org>
Sat, 7 Nov 2015 10:18:38 +0000 (10:18 +0000)
commit0aad01985747ab503fd71ccf4767d0069fc3e85a
treeccf3e9b48b09616d6c7845a4b402ae67fa434d63
parent46e343b8d225732ae2e81509ca1accf47444f64b
Move #undef DEF_BUILTIN* to builtins.def

I was confused at first why tree-core.h was undefining DEF_BUILTIN_CHKP
before defining it, then undefining it again after including builtins.def.
This is because builtins.def provides a default definition of
DEF_BUILTIN_CHKP, but leaves it up to the caller to undefine it where
necessary.  Similarly to the previous internal-fn.def patch, it seems
more obvious for builtins.def to #undef things unconditionally.

One argument might have been that keeping preprocessor stuff
out of the .def files makes it easier for non-cpp parsers.  In practice
though we already have #ifs and multiline #defines, so single-line #undefs
should be easy in comparison.

Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.

gcc/ada/
* gcc-interface/utils.c: Don't undef DEF_BUILTIN.

gcc/c-family/
* c-common.c: Don't undef DEF_BUILTIN.

gcc/jit/
* jit-builtins.c: Don't undef DEF_BUILTIN.

gcc/lto/
* lto-lang.c: Don't undef DEF_BUILTIN.

gcc/
* builtins.def: #undef DEF_BUILTIN and DEF_BUILTIN_CHKP
* builtins.c, genmatch.c, tree-core.h: Don't undef them here.

From-SVN: r229924
13 files changed:
gcc/ChangeLog
gcc/ada/ChangeLog
gcc/ada/gcc-interface/utils.c
gcc/builtins.c
gcc/builtins.def
gcc/c-family/ChangeLog
gcc/c-family/c-common.c
gcc/genmatch.c
gcc/jit/ChangeLog
gcc/jit/jit-builtins.c
gcc/lto/ChangeLog
gcc/lto/lto-lang.c
gcc/tree-core.h