[v4,1/2] libkrb5: New package

Signed-off-by: André Hentschel <nerv@dawncrow.de>
---
 DEVELOPERS                   |  1 +
 package/Config.in            |  1 +
 package/libkrb5/Config.in    | 11 +++++++++++
 package/libkrb5/libkrb5.hash |  2 ++
 package/libkrb5/libkrb5.mk   | 22 ++++++++++++++++++++++
 5 files changed, 37 insertions(+)
 create mode 100644 package/libkrb5/Config.in
 create mode 100644 package/libkrb5/libkrb5.hash
 create mode 100644 package/libkrb5/libkrb5.mk

Hello,

Thanks for this new iteration. Much better without the
configure.ac patch.

On Mon, 20 Nov 2017 22:32:28 +0100, André Hentschel wrote:

> +LIBKRB5_VERSION = 1.15.2
> +LIBKRB5_SITE = https://web.mit.edu/kerberos/dist/krb5/1.15
> +LIBKRB5_SOURCE = krb5-$(LIBKRB5_VERSION).tar.gz
> +LIBKRB5_SUBDIR = src
> +LIBKRB5_LICENSE = MIT
> +LIBKRB5_LICENSE_FILES = NOTICE
> +LIBKRB5_INSTALL_STAGING = YES
> +LIBKRB5_AUTORECONF = YES

Are you sure you still need to AUTORECONF the package? Now that the
configure.ac patch is gone, I don't think this is necessary anymore.

> +# Add sane assumptions about cross-compiling

Rather than "sane assumptions", a clearer comment would be: "The
configure script uses AC_TRY_RUN tests to check for those values, which
doesn't work in a cross-compilation scenario. Therefore, we feed the
configure script with the correct answer for those tests". Or something
along those lines.

Note: both issues can be fixed when applying, no need to resend just
for that.

Thanks!

Thomas

Am 21.11.2017 um 13:56 schrieb Thomas Petazzoni:
> Hello,
> 
> Thanks for this new iteration. Much better without the
> configure.ac patch.
> 
> On Mon, 20 Nov 2017 22:32:28 +0100, André Hentschel wrote:
> 
>> +LIBKRB5_VERSION = 1.15.2
>> +LIBKRB5_SITE = https://web.mit.edu/kerberos/dist/krb5/1.15
>> +LIBKRB5_SOURCE = krb5-$(LIBKRB5_VERSION).tar.gz
>> +LIBKRB5_SUBDIR = src
>> +LIBKRB5_LICENSE = MIT
>> +LIBKRB5_LICENSE_FILES = NOTICE
>> +LIBKRB5_INSTALL_STAGING = YES
>> +LIBKRB5_AUTORECONF = YES
> 
> Are you sure you still need to AUTORECONF the package? Now that the
> configure.ac patch is gone, I don't think this is necessary anymore.
> 
>> +# Add sane assumptions about cross-compiling
> 
> Rather than "sane assumptions", a clearer comment would be: "The
> configure script uses AC_TRY_RUN tests to check for those values, which
> doesn't work in a cross-compilation scenario. Therefore, we feed the
> configure script with the correct answer for those tests". Or something
> along those lines.
> 
> Note: both issues can be fixed when applying, no need to resend just
> for that.

Yes, thanks for taking care!
The autoreconf isn't needed anymore, I just tested it.

Hello,

On Mon, 20 Nov 2017 22:32:28 +0100, André Hentschel wrote:
> Signed-off-by: André Hentschel <nerv@dawncrow.de>
> ---
>  DEVELOPERS                   |  1 +
>  package/Config.in            |  1 +
>  package/libkrb5/Config.in    | 11 +++++++++++
>  package/libkrb5/libkrb5.hash |  2 ++
>  package/libkrb5/libkrb5.mk   | 22 ++++++++++++++++++++++
>  5 files changed, 37 insertions(+)
>  create mode 100644 package/libkrb5/Config.in
>  create mode 100644 package/libkrb5/libkrb5.hash
>  create mode 100644 package/libkrb5/libkrb5.mk

I'm sorry but this doesn't build. First, I get:

prof_file.c: In function ‘profile_update_file_data_locked’:
prof_file.c:370:16: error: ‘f’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
         retval = profile_parse_file(f, &data->root, ret_modspec);

This is because -Werror=uninitialized. Looking at the code, I believe
gcc is wrong. So either we should initialize f = NULL at the beginning
of the function, or we should remove -Werror=uninitialized from the
build flags. After fixing this, I have a second build failure:

In file included from net-server.c:61:0:
../../include/net-server.h:32:19: fatal error: verto.h: No such file or directory
 #include <verto.h>
                   ^
compilation terminated.

You can reproduce by building the following defconfig:

BR2_arm=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_TOOLCHAIN_EXTERNAL_CUSTOM=y
BR2_TOOLCHAIN_EXTERNAL_DOWNLOAD=y
BR2_TOOLCHAIN_EXTERNAL_URL="http://autobuild.buildroot.org/toolchains/tarballs/br-arm-full-2017.11-rc1.tar.bz2"
BR2_TOOLCHAIN_EXTERNAL_GCC_4_9=y
BR2_TOOLCHAIN_EXTERNAL_HEADERS_3_10=y
BR2_TOOLCHAIN_EXTERNAL_LOCALE=y
# BR2_TOOLCHAIN_EXTERNAL_HAS_THREADS_DEBUG is not set
BR2_TOOLCHAIN_EXTERNAL_CXX=y
BR2_INIT_NONE=y
BR2_SYSTEM_BIN_SH_NONE=y
# BR2_PACKAGE_BUSYBOX is not set
BR2_PACKAGE_LIBKRB5=y
# BR2_TARGET_ROOTFS_TAR is not set

Please use the test-pkg script to build test your package for the next
submission (and don't forget to remove AUTORECONF = YES, and to adjust
the comment about CONF_ENV, as we discussed previously).

Thanks a lot!

Thomas