/u/sy/beebe/tex/ndvi/cxref/CHANGES, Wed Mar 21 15:25:34 1990
Edit by Nelson H.F. Beebe <beebe@plot79.math.utah.edu>

	     Change Log for Arnold Robbins' cxref Program

[23-Mar-1990]
		Update for full ANSI C support:

		(1) Add recognition of new ANSI keywords `noalias' and
		`signed'; the other new ones, `const', `enum', `void',
		and `volatile', were already supported.

		(2) Add recognition of hexadecimal escape sequences in
		character and string constants.

		(3) Add recognition of  integer (l, L, u,  U) constant
		suffixes, and floating (f, F, l, F) suffixes.

		(4) Add recognition  of   wide character (L'...')  and
		string (L"...") constants.

		(5) Add recognition of ANSI trigraphs: 3-character
		(??x) representations of # [ \ ] ^ { | } ~.

		Support for the following common C language extensions
		has also been added:

		(1) $ in names (VAX and HP/Apollo)

		(2)  IBM PC DOS   (Borland Turbo, Lattice,  Microsoft,
		TopSpeed)  type  modifiers  `cdecl', `far', `fortran',
		`huge', `interrupt', `near', and `pascal'.

		(3) GNU  C (gcc),  and  a  few others,   `inline' type
		modifier.

		"make docxref" will run a  test  of the parser using a
		special file,  test.c; the  output should be identical
		to test.ok.

[21-Mar-1990]
		Legal C input like

			putchar('\"');

		confused the scanner.  If   there was  only one   such
		instance,  an error message about  unexpected  EOF was
		issued.   If there were  two, then  some   of the text
		between was  output as a  string,  and garbage  tokens
		were produced (by lex's automatic echoing of things it
		doesn't understand).

		The  problem was that  the character  constant grammar
		rule pattern in  cscan.l did not correctly handle this
		case.  Here are the changes:

% diff cscan.l-org cscan.l
94a95
>
97c98
< '\\[\\bfrnlt']'                       outchar();
---
> '\\[\\abfnrtv"?']'            outchar();

		Note that  \a, \", \v,  and \? are  now recognized, as
		per ANSI C.  \| is no  longer recognized (it has never
		been a legal C escape).