Well, I don't really feel qualified to comment since I'm unfamiliar with
gcc internals, but since nobody else seems to be following your
progress, I'll throw in my US$0.02
This seems like a simple fix and I can't think of any other test cases
that you haven't already addressed. I say we add it to the "good patch"
pile. I don't recall if this came up because of gcc test base analysis
or from an avr-gcc user's bug, but I suppose we would want a regression
test if we don't have one.
Thanks for you efforts!
-dave
Andy H wrote:
I tested it with PROGMEM and it returns byte address (same as now)
Tested it with Labels ( ptr = &&foo) and it return word address (which I
intended).
I did not test goto *ptr, as I have not loaded patch for jump/long jump
(so its does nothing). I'm sure my patch for that expects words so it
will not be problem.
I have checked switch/jump tables. That is un-changed using WORDs,
To Summarize:
Address of function WORDs
Address of function +- offset WORDS with WORD offset
Address of normal RAM variable BYTEs
Address of PROGMEM variable BYTEs
Address of Label WORDS
Address used in switch table WORDs
Pointer used in goto *ptr WORDs (TBC)
Looks good so far!
Comments?
_______________________________________________
AVR-GCC-list mailing list
AVR-GCC-list@nongnu.org
http://lists.nongnu.org/mailman/listinfo/avr-gcc-list