Re: [OE-core] [PATCH v4] ltp: add support for tirpc and fix build issue

["Richard Purdie" <richard.purdie@linuxfoundation.org>]

On Sat, 2020-12-19 at 12:08 +0100, Kory Maincent wrote:
> Add support to tirpc tests build.
> The export of ${prefix} updates all other paths variable and
> in particular PKG_CONFIG_PATH. This variable beeing updated it could
> not find the pkgconfig file for tirpc.
> Use immediate expansion to avoid ${prefix} export overflows on
> PKG_CONFIG_PATH.
> 
> Signed-off-by: Kory Maincent <kory.maincent@bootlin.com>
> ---
> 
> Change since v1:
>  - Use PACKAGECONFIG to enable rpc tests
>  - Improve ${libdir} and ${incdir} export
> 
> Change since v2:
>  - Improve ${libdir} and ${incdir} export
> 
> Change since v3:
>  - Use PKG_CONFIG_PATH in place of the ${libdir} ${incdir} variable export
> 
>  meta/recipes-extended/ltp/ltp_20200930.bb | 5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
> 
> diff --git a/meta/recipes-extended/ltp/ltp_20200930.bb b/meta/recipes-extended/ltp/ltp_20200930.bb
> index 7acf15b360..1a9c05a7e4 100644
> --- a/meta/recipes-extended/ltp/ltp_20200930.bb
> +++ b/meta/recipes-extended/ltp/ltp_20200930.bb
> @@ -42,14 +42,13 @@ inherit autotools-brokensep pkgconfig
>  
>  TARGET_CC_ARCH += "${LDFLAGS}"
>  
> +PKG_CONFIG_PATH := "${PKG_CONFIG_PATH}"
>  export prefix = "/opt/${PN}"
> -export exec_prefix = "/opt/${PN}"
>  
>  PACKAGECONFIG[numa] = "--with-numa, --without-numa, numactl,"
> +PACKAGECONFIG[rpc] = "--with-tirpc, --without-tirpc, virtual/librpc,"
>  EXTRA_AUTORECONF += "-I ${S}/testcases/realtime/m4"
>  EXTRA_OECONF = " --with-realtime-testsuite --with-open-posix-testsuite "
> -# ltp network/rpc test cases ftbfs when libtirpc is found
> -EXTRA_OECONF += " --without-tirpc "

I did queue this for testing however we saw:

https://autobuilder.yoctoproject.org/typhoon/#/builders/86/builds/1639/steps/14/logs/stdio

(and I suspect we'll see the other selftests fail too).

I suspect this is due to hardcoded paths being expanded into
PKG_CONFIG_PATH which makes the recipe's sstate depend upon the
directory it was built within which is a real problem we need to fix.

Cheers,

Richard