Subject: re: Recap on PR 29689
To: segv <segv@netctl.net>
From: matthew green <mrg@eterna.com.au>
List: port-sparc64
Date: 08/18/2005 11:24:09
On Thu, 18 Aug 2005 00:48:57 +0100
segv <segv@netctl.net> wrote:
> sparc.o.tmp /opt/xsrc/xfree/xc/extras/Mesa/src/mesa/sparc/sparc.c: In function
> `_mesa_init_sparc_glapi_relocs': /opt/xsrc/xfree/xc/extras/Mesa/src/mesa/sparc/sparc.c:167:
> warning: right shift count >= width of
> type /opt/xsrc/xfree/xc/extras/Mesa/src/mesa/sparc/sparc.c:170: warning: right
> shift count >= width of type *** [sparc.o] Error code 1 1 error
OK, looking at sparc.c file, it seems that '__sparc_v9__' macro gets defined,
when it shouldn't, so it assumes 'unsigned long' are 64-bit datatypes
well, __sparc_v9__ macro means "sparc64 cpu" not "64 bit code". that
what __arch64__ is for. __sparc_v9__ means you can use v9 instruction.
i guess the code needs to be fixed.
What I don't understand is how passing '-mcpu=ultrasparc' to gcc, triggers the
above macro being defined??
PS The build was run with './build.sh -m sparc ...'