Re: [PATCH 3/3] texinfo: fix info not work

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

On Tue, 2014-09-23 at 22:19 +0800, Hongxu Jia wrote:
> In texinfo, we didn't have info/dir entries, which caused
> info not work.
> ...
> root@qemux86:~# info info
> info: dir: No such file or directory
> ...
> Invoke install-info to create info/dir entries in pkg_postinst_info.
> 
> Also fix a QA Issue
> ...
> ERROR: QA Issue: texinfo: Files/directories were installed but not shipped
>   /usr/share/info/info.info.bz2
>   /usr/share/info/info-stnd.info.bz2 [installed-vs-shipped]
> ...
> 
> [YOCTO #6751]
> 
> Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
> ---
>  meta/recipes-extended/texinfo/texinfo_5.2.bb | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/meta/recipes-extended/texinfo/texinfo_5.2.bb b/meta/recipes-extended/texinfo/texinfo_5.2.bb
> index be90aba..91552a9 100644
> --- a/meta/recipes-extended/texinfo/texinfo_5.2.bb
> +++ b/meta/recipes-extended/texinfo/texinfo_5.2.bb
> @@ -51,7 +51,7 @@ do_install_append_class-native() {
>  PACKAGES += "info info-doc"
>  
>  FILES_info = "${bindir}/info ${bindir}/infokey ${bindir}/install-info"
> -FILES_info-doc = "${infodir}/info.info ${infodir}/dir ${infodir}/info-*.info \
> +FILES_info-doc = "${infodir}/info.info* ${infodir}/dir ${infodir}/info-*.info* \
>                    ${mandir}/man1/info.1* ${mandir}/man5/info.5* \
>                    ${mandir}/man1/infokey.1* ${mandir}/man1/install-info.1*"
>  
> @@ -61,4 +61,10 @@ FILES_${PN}-doc = "${infodir}/texinfo* \
>                     ${datadir}/${tex_texinfo} \
>                     ${mandir}/man1 ${mandir}/man5"
>  
> +pkg_postinst_info () {
> +    for f in `find $D/${infodir} -name "*.info*"`; do
> +        install-info $f $D/${infodir}/dir
> +    done
> +}
> +
>  BBCLASSEXTEND = "native"


I think there are a few issues here. Firstly, this postinst can run at
build time which means there is a dependency on texinfo-native. I guess
if this fails, it will just run on target so its probably "harmless" but
not deterministic. Secondly, anything installing an info directory
really needs to update the dir files with its own postinst. For optimal
handling at rootfs time, that will need trigger/intercept support.

So this probably can merge as it improves things however we need some
enhacement bugs opening related to this, we may want to force the
postinst to run on target for now so things are deterministic too.

Cheers,

Richard