[bitc-dev] bug in character decode routines

[Sam Mason]

Hi,

Another little buglet -- this time in the character decoding routines!
If you enter a character literal like #\a it will be decoded as some
random value (``codePoint'' in LitValue::DecodeRawCharacter() never
appears to get initialised to anything in this case).

As far as I can tell (a couple of the unit tests seem to confirm it)
this is how I would write the single character `a' and I'm not expected
to write #\{U+61}.  Even if you are, I don't think the computer should
be writing out random values from uninitialised variables.


  Sam
-------------- next part --------------
--- src.coyotos/src/ccs/bitcc-bootstrap/frontend/LitValue.cxx	2006-05-15 19:13:06.000000000 +0100
+++ src.coyotos/src/ccs/bitcc-bootstrap/frontend/LitValue.cxx	2006-05-15 19:13:07.000000000 +0100
@@ -267,6 +267,10 @@
     else if (strcmp(s, "rbrace") == 0) {
       codePoint = '}';
       s += 6;
+    } else {
+      /* finally, if it's not one of those, just treat it as a single
+       * character */
+      codePoint = *s++;
     }
   }