Re: gnumach FTBFS

[Svante Signell]

On Wed, 2011-05-18 at 15:25 +0200, Svante Signell wrote:
> On Wed, 2011-05-18 at 13:35 +0200, Svante Signell wrote:
> > On Wed, 2011-05-18 at 12:18 +0200, Thomas Schwinge wrote:
> ...
> > I have now completely disabled the wrapper, and the build problems
> > remain.
> 
> And building with -O0 does not work either.

Looks like at least -O1 is needed: It builds OK at that level, but still
not at -O2!

A question: I can avoid the warnings below by doing the following asm
change. I saw it somewhere on the web that people recommended this type
of change to enable the inline code go into a register. Does it make
sense (I don't remember much of asm)

  asm("xorl %0,%0\n0:\tbsfw %1,%w0\n\tbtr %0,%1\n\tjnc 0b"
-      : "=&r" (rv), "=m" (*field) : "1" (*field));
+      : "=&r" (rv), "=m" (*field) : "m" (*field));

../linux/src/drivers/scsi/ultrastor.c: In function
‘find_and_clear_bit_16’:
../linux/src/drivers/scsi/ultrastor.c:308:3: warning: matching
constraint does not allow a register [enabled by default]
(... same several times ...)

-> The error below goes away with -O1 or higher!
-> ../linux/src/drivers/scsi/ultrastor.c:308:3: error: inconsistent
operand constraints in an ‘asm’

Is this OK? There is also a similar case in the glibc code.