Re: [JAVA] Avoid use of CHAR_TYPE tree codes.

[Andrew Haley <aph at redhat dot com>]

Roger Sayle writes:
 > 
 > On 27 Jan 2006, Tom Tromey wrote:
 > > The last time removing CHAR_TYPE came up, Jim Wilson pointed out that
 > > this affects debug information..
 > 
 > Now that the handling of dwarf2 debug information for character types
 > has been cleaned up in the middle-end, there's no longer a reason for
 > the Java front-end to use CHAR_TYPE nodes, as the same information can
 > be represented using TYPE_STRING_FLAG on INTEGER_TYPE nodes.
 > 
 > The following patch is a repost of the original to eradicate the use
 > of CHAR_TYPE from the Java front-end, updated to set and preserve
 > TYPE_STRING_FLAG appropriately.  For testing to show why this is
 > safe see: http://gcc.gnu.org/ml/gcc-patches/2006-01/msg01974.html
 > 
 > Tested on x86_64-unknown-linux-gnu with a full "make bootstrap",
 > including java, and regression tested with a top-level "make -k check"
 > with no new failures.
 > 
 > Ok for mainline?
 > 
 > 
 > 
 > 2006-02-04  Roger Sayle  <roger@eyesopen.com>
 > 
 > 	* decl.c (java_init_decl_processing): Create char_type_node as a
 > 	regular INTEGER_TYPE node.
 > 	(push_promoted_type): Preserve TYPE_STRING_FLAG on types.
 > 	* typeck.c (convert): No longer check for CHAR_TYPEs but instead
 > 	test for char_type_node and promoted_char_type_node as special
 > 	instances of INTEGER_TYPE tree codes.
 > 	(promote_type,build_java_signature): Likewise.
 > 	* jcf-write.c (adjust_typed_op): Likewise.
 > 	* mangle.c (mangle_type): Likewise.
 > 	* parse.y (do_unary_numeric_promotion): No longer handle CHAR_TYPE.
 > 	* parse.h (JINTEGRAL_TYPE_P): Likewise.

Yes, thanks.  This change is quite delicate becasue it affects the
mangling, but AFAICS you've do the right thing.

Andrew.