Re: [flashrom] compiling flashrom for Debian/kFreeBSD

[Carl-Daniel Hailfinger]

Hi Robert,

thanks for testing that patch!

Am 24.02.2012 20:14 schrieb Robert Millan:
> Long time no see :-)

Indeed.
> El 24 de febrer de 2012 0:55, Carl-Daniel Hailfinger
>  ha escrit:
> > I have written a patch which should fix compilation and
> > hopefully also work during runtime.
>
> The patch works perfectly. However, some changes are gratuitous or inaccurate:
> > -#elif defined(__FreeBSD__) || defined (__DragonFly__)
> > +#elif defined(__FreeBSD__) || defined(__FreeBSD_kernel__) || defined (__DragonFly__)
> >    if ((io_fd = open("/dev/io", O_RDWR)) < 0) {
> >  #else
> >    if (iopl(3) != 0) {
>
> We have both interfaces. You can keep using iopl() if you like.

Hm. Is iopl(3) a feature present in the FreeBSd kernel or is it just
part of the compat layer presented by glibc?
> > #if defined(__FreeBSD__) || defined(__DragonFly__)
> > /* Note that FreeBSD kernel with glibc does not have machine/cpufunc.h. */
> > #include <machine/cpufunc.h>
> > #define off64_t off_t
> > #define lseek64 lseek
>
> Actually, we have machine/cpufunc.h (we have everything in machine/*
> actually).  The off_t / lseek defines after this look dangerous though
> (they collide with glibc definitions).

Comment has been replaced with
/* Note that FreeBSD kernel with glibc has conflicting out[bwl] definitions
 * in machine/cpufunc.h and sys/io.h. Use cpufunc.h only for plain FreeBSD.
 */
> > A full build log would be nice to have so I can see any remaining build
> > issues.
>
> Attached (build.log).

Looks good, and works as intended.
> > If you could test if flashrom works (do NOT test flashrom on
> > laptops/notebooks/servers), that would be even better. Just run
> >
> > flashrom -VV
>
> Attached (run.log).

The SFDP limitation you were hitting has been fixed a few minutes ago in
r1506 in flashrom svn.

If you have any other comments about my changes, please tell me.

Regards,
Carl-Daniel

compiling flashrom for Debian/kFreeBSD

[Robert Millan]

El 25 de febrer de 2012 0:49, Carl-Daniel Hailfinger >> El 24 de
febrer de 2012 0:55, Carl-Daniel Hailfinger
> >  ha escrit:
> > We have both interfaces. You can keep using iopl() if you like.
>
> Hm. Is iopl(3) a feature present in the FreeBSd kernel or is it just
> part of the compat layer presented by glibc?

It's a library function (glibc). Just pick the one you prefer :-)
> > > #include <machine/cpufunc.h>
> > >    #define off64_t off_t
> > >    #define lseek64 lseek
> >
> > Actually, we have machine/cpufunc.h (we have everything in machine/*
> > actually).  The off_t / lseek defines after this look dangerous though
> > (they collide with glibc definitions).
>
> Comment has been replaced with
> /* Note that FreeBSD kernel with glibc has conflicting out[bwl] definitions
>  * in machine/cpufunc.h and sys/io.h. Use cpufunc.h only for plain FreeBSD.
>  */

Uhm no, we removed the conflicts.  We need to allow programs to
include both headers painlessly.  Again, just pick the one you prefer.
 cpufunc.h is supposed to work, and if it doesn't it's a bug.
> > > #define off64_t off_t

Needed because FreeBSD lacks off64_t. We have both off64_t and off_t
(both are 64-bit) so on GNU/kFreeBSD this define is harmless but not
useful.
> > > #define lseek64 lseek

Same here.