[Buildroot] [PATCH v2] package/nfs-utils: improve local patch

[Giulio Benetti]

Hi Thomas,

On 11/08/22 22:47, Thomas Petazzoni via buildroot wrote:
> On Thu, 11 Aug 2022 01:38:24 +0200
> Giulio Benetti <giulio.benetti at benettiengineering.com> wrote:
> 
>> As Thomas Petazzoni made me notice the actual local patch is not a good way
>> at all to deal with pkg-config with Autotools. So let's rework the patch
> 
> It is recommended to avoid first person sentences in commit logs.
> 
>> according to his comment that represents the standard of Autotools
>> pkg-config usage. This impacts host-nfs-utils too that is used to build
>> nfs-utils and so libmount must be present in host. We do this by adding
>> host-util-linux to HOST_NFS_UTILS_DEPENDENCIES since it is the libmount
>> provider.
> 
> So, host-utils-linux is now only needed by host-nfs-utils because of
> the new check in configure.ac, but nfsrahead is in fact not built for the host?
> 
> If it's the case, then the configure.ac change is wrong, and the mount
> library of util-linux should only be checked using PKG_CHECK_MODULES()
> when it is actually needed.

Totally, and I've managed to work this recursive issue with a patch I'm
going to send soon.

> Another nfs-utils thing I noticed:
> 
> ifeq ($(BR2_PACKAGE_UTIL_LINUX_LIBBLKID),y)
> NFS_UTILS_CONF_OPTS += --enable-uuid
> else
> NFS_UTILS_CONF_OPTS += --disable-uuid
> endif
> 
> but BR2_PACKAGE_NFS_UTILS selects BR2_PACKAGE_UTIL_LINUX_LIBBLKID so
> this conditional doesn't make much sense.

I've removed those lines in the same since it's a left-over of a
previous patch and libuuid is needed by default now.

Thank you for pointing me the lines above, this was a pretty difficult
patch for me :-) I hope the new patch will be correct!

Best regards
-- 
Giulio Benetti
Benetti Engineering sas