Re: [PATCH] build: Use AC_HEADER_MAJOR to find device macros

[Andrei Borzenkov <arvidjaar@gmail.com>]

17.04.2016 18:28, Mike Gilbert пишет:
> On Sun, Apr 17, 2016 at 1:53 AM, Andrei Borzenkov <arvidjaar@gmail.com> wrote:
>> 17.04.2016 00:34, Mike Gilbert пишет:
>>> Depending on the OS/libc, device macros may be found in 3 places:
>>>
>>
>> Mentioning OS and libc versions that have problem would be helpful.
>>
> 
> I am really only familiar with glibc, though I believe BSD and Sun use
> sys/mkdev.h?
> 
>>> sys/types.h
>>> sys/mkdev.h
>>> sys/sysmacros.h
>>>
>>> glibc currenctly defines the major/minor/makedev macros in sys/sysmacros.h
>>> and includes this from sys/types.h. Based on mailing list discussion,
>>> this may be removed from sys/types.h in a future glibc release.
>>> ---
>>>  configure.ac                         | 3 ++-
>>>  grub-core/osdep/devmapper/getroot.c  | 6 ++++++
>>>  grub-core/osdep/devmapper/hostdisk.c | 5 +++++
>>>  grub-core/osdep/linux/getroot.c      | 6 ++++++
>>
>> All those files are linux only. Do you mean there are some linux flavors
>> that have these definitions in non-standard place? Could you name them?
>>
> 
> As I mentioned, in a future release, glibc may remove #include
> <sys/sysmacros.h> from sys/types.h. This would provide preventing
> naming conflicts in programs that have variables named like "major"
> and indirectly include sys/types.h. This is currently up for
> discussion upstream.
> 
> https://sourceware.org/ml/libc-alpha/2015-11/msg00253.html
> 
> https://sourceware.org/ml/libc-alpha/2015-12/msg00612.html
> 

Then your commit message is misleading. It is not found in 3 places, it
is found in 2 places, one of which is pulled in by default currently.

> If you would prefer to wait until a commit lands in glibc, I
> understand. We have been doing some testing in Gentoo to see how many
> things will break (it's a lot!). I just thought I would submit this
> now since the check is fairly harmless and would future-proof the code
> either way.

I'm fine with it actually (it should not break anything) but could you
please clean up commit message, mention that problem will be caused by
removal of sys/sysmacros.h default inclusion in glibc and add reference
to above discussion. Given release cycle of grub it probably make sense
to include it in 2.02.

> 
>>>  grub-core/osdep/unix/getroot.c       | 4 +++-
>>>  5 files changed, 22 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/configure.ac b/configure.ac
>>> index 57e1713..9ddfc53 100644
>>> --- a/configure.ac
>>> +++ b/configure.ac
>>> @@ -388,7 +388,8 @@ fi
>>>
>>>  # Check for functions and headers.
>>>  AC_CHECK_FUNCS(posix_memalign memalign getextmntent)
>>> -AC_CHECK_HEADERS(sys/param.h sys/mount.h sys/mnttab.h sys/mkdev.h limits.h)
>>> +AC_CHECK_HEADERS(sys/param.h sys/mount.h sys/mnttab.h limits.h)
>>> +AC_HEADER_MAJOR
>>>
>>>  AC_CHECK_MEMBERS([struct statfs.f_fstypename],,,[$ac_includes_default
>>>  #include <sys/param.h>
>>> diff --git a/grub-core/osdep/devmapper/getroot.c b/grub-core/osdep/devmapper/getroot.c
>>> index 05eda50..72e5582 100644
>>> --- a/grub-core/osdep/devmapper/getroot.c
>>> +++ b/grub-core/osdep/devmapper/getroot.c
>>> @@ -40,6 +40,12 @@
>>>  #include <limits.h>
>>>  #endif
>>>
>>> +#if defined(MAJOR_IN_MKDEV)
>>> +#include <sys/mkdev.h>
>>> +#elif defined(MAJOR_IN_SYSMACROS)
>>> +#include <sys/sysmacros.h>
>>> +#endif
>>> +
>>>  #include <libdevmapper.h>
>>>
>>>  #include <grub/types.h>
>>> diff --git a/grub-core/osdep/devmapper/hostdisk.c b/grub-core/osdep/devmapper/hostdisk.c
>>> index 19c1101..a697bcb 100644
>>> --- a/grub-core/osdep/devmapper/hostdisk.c
>>> +++ b/grub-core/osdep/devmapper/hostdisk.c
>>> @@ -24,6 +24,11 @@
>>>  #include <errno.h>
>>>  #include <limits.h>
>>>
>>> +#if defined(MAJOR_IN_MKDEV)
>>> +#include <sys/mkdev.h>
>>> +#elif defined(MAJOR_IN_SYSMACROS)
>>> +#include <sys/sysmacros.h>
>>> +#endif
>>>
>>>  #ifdef HAVE_DEVICE_MAPPER
>>>  # include <libdevmapper.h>
>>> diff --git a/grub-core/osdep/linux/getroot.c b/grub-core/osdep/linux/getroot.c
>>> index 10480b6..09e7e6e 100644
>>> --- a/grub-core/osdep/linux/getroot.c
>>> +++ b/grub-core/osdep/linux/getroot.c
>>> @@ -35,6 +35,12 @@
>>>  #include <limits.h>
>>>  #endif
>>>
>>> +#if defined(MAJOR_IN_MKDEV)
>>> +#include <sys/mkdev.h>
>>> +#elif defined(MAJOR_IN_SYSMACROS)
>>> +#include <sys/sysmacros.h>
>>> +#endif
>>> +
>>>  #include <grub/types.h>
>>>  #include <sys/ioctl.h>         /* ioctl */
>>>  #include <sys/mount.h>
>>> diff --git a/grub-core/osdep/unix/getroot.c b/grub-core/osdep/unix/getroot.c
>>> index 1079a91..4bf37b0 100644
>>> --- a/grub-core/osdep/unix/getroot.c
>>> +++ b/grub-core/osdep/unix/getroot.c
>>> @@ -51,8 +51,10 @@
>>>  #endif
>>>
>>>  #include <sys/types.h>
>>> -#if defined(HAVE_SYS_MKDEV_H)
>>> +#if defined(MAJOR_IN_MKDEV)
>>>  #include <sys/mkdev.h>
>>> +#elif defined(MAJOR_IN_SYSMACROS)
>>> +#include <sys/sysmacros.h>
>>>  #endif
>>>
>>
>> The names are really misleading. All that this macro checks for is
>> whether these headers are present, so it is entirely equivalent to
>> AC_CHECK_HEADERS([sys/mkdev.h sys/sysmacros.h]). Which returns us to the
>> question which systems have sys/sysmacros.h :)
>>
> 
> The macro has existed for over 20 years. I can't say why it was named that way.
> 
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel
>