Re: [coreboot] QEMU build failure in master

[Timothy Pearson]

On 05/06/2015 11:46 AM, Aaron Durbin wrote:
> On Wed, May 6, 2015 at 9:45 AM, Timothy Pearson
>   wrote:
> > On 05/06/2015 11:41 AM, Aaron Durbin wrote:
> > > That's probably my fault. I was under the impression monotonic_timer
> > > was a first class citizen now (I at least recall someone doing that) I
> > > thought wrong?
> > >
> > > You could add the following in the beginning of that function:
> > >
> > > if (!IS_ENABLED(CONFIG_HAVE_MONOTONIC_TIMER)
> > >      return 0;
> > >
> > > I need to look at your logs to know the build failure.
> >
> > While I don't have the entire log available right now, this was the
> > important part:
> > coreboot/src/lib/timestamp.c:184: undefined reference to
> > `timer_monotonic_get'
> >
> > There were no other errors, and no other warning output except from iasl (as
> > usual).
> >
> > It looks like timer_monotonic_get may not be available on QEMU, which means
> > your fix wouldn't work unless the entire block was #ifdefed out.
>
> The compiler would throw out that code not used using dead code elimination.

All right, I wasn't sure if we wanted to rely on that or not.  It 
"feels" fragile at first glance, and might provide an unpleasant 
surprise down the line to a programmer less familiar with the internal 
workings of compilers.

QEMU build failure in master

[Aaron Durbin]

On Wed, May 6, 2015 at 9:51 AM, Timothy Pearson
 wrote:
> On 05/06/2015 11:46 AM, Aaron Durbin wrote:
> > On Wed, May 6, 2015 at 9:45 AM, Timothy Pearson
> >   wrote:
> > > On 05/06/2015 11:41 AM, Aaron Durbin wrote:
> > > > That's probably my fault. I was under the impression monotonic_timer
> > > > was a first class citizen now (I at least recall someone doing that) I
> > > > thought wrong?
> > > >
> > > > You could add the following in the beginning of that function:
> > > >
> > > > if (!IS_ENABLED(CONFIG_HAVE_MONOTONIC_TIMER)
> > > >      return 0;
> > > >
> > > > I need to look at your logs to know the build failure.
> > >
> > > While I don't have the entire log available right now, this was the
> > > important part:
> > > coreboot/src/lib/timestamp.c:184: undefined reference to
> > > `timer_monotonic_get'
> > >
> > > There were no other errors, and no other warning output except from iasl
> > > (as
> > > usual).
> > >
> > > It looks like timer_monotonic_get may not be available on QEMU, which
> > > means
> > > your fix wouldn't work unless the entire block was #ifdefed out.
> >
> > The compiler would throw out that code not used using dead code
> > elimination.
>
> All right, I wasn't sure if we wanted to rely on that or not.  It "feels"
> fragile at first glance, and might provide an unpleasant surprise down the
> line to a programmer less familiar with the internal workings of compilers.

The upside is that all he code paths are checked by the compiler
instead of going stale if the c preprocessor throws out the code.
>