Re: [ptxdist] [PATCH 4/4] ptxd_make_world_inject: Introduce separate destination directory

[Alexander Dahl <ada@thorsis.com>]

Hello Michael,

thanks for reviewing, answers below …

Am Fri, May 03, 2024 at 11:48:52AM +0200 schrieb Michael Olbrich:
> On Wed, Apr 24, 2024 at 04:31:09PM +0200, Alexander Dahl wrote:
> > Setting the optional <PKG>_INJECT_DEST now allows to give a different
> > target for putting the binary blobs into.  When building out-of-tree
> > some bootloaders like U-Boot expect injected files in the build dir, not
> > in the source dir.  Backwards compatibility is ensured, the source dir
> > is still the default, the new inject dest is optional and in case set it
> > overwrites the default.
> > 
> > For using this in u-boot package, on top of the things already done to
> > use the inject mechanism in general, add this line to
> > 'rules/u-boot.make' for example:
> > 
> >     U_BOOT_INJECT_DEST     := $(U_BOOT_BUILD_DIR)
> 
> Can you add a patch for u-boot?

I will, in an upcoming patch series with more changes to the u-boot
package.  Planned to this after lunch later today anyways. ^^

> > 
> > Signed-off-by: Alexander Dahl <ada@thorsis.com>
> > ---
> >  rules/post/ptxd_make_world_inject.make |  3 ++-
> >  scripts/lib/ptxd_make_world_inject.sh  | 22 ++++++++++++++++++----
> >  2 files changed, 20 insertions(+), 5 deletions(-)
> > 
> > diff --git a/rules/post/ptxd_make_world_inject.make b/rules/post/ptxd_make_world_inject.make
> > index eabcdd052..509d480ba 100644
> > --- a/rules/post/ptxd_make_world_inject.make
> > +++ b/rules/post/ptxd_make_world_inject.make
> > @@ -9,7 +9,8 @@
> >  world/inject/env = \
> >  	$(call world/env, $(1)) \
> >  	pkg_inject_path="$(call ptx/escape,$($(1)_INJECT_PATH))" \
> > -	pkg_inject_files="$(call ptx/escape,$($(1)_INJECT_FILES))"
> > +	pkg_inject_files="$(call ptx/escape,$($(1)_INJECT_FILES))" \
> > +	pkg_inject_dest="$(call ptx/escape,$($(1)_INJECT_DEST))"
> >  
> >  world/inject = \
> >  	$(call world/inject/env,$(strip $(1))) \
> > diff --git a/scripts/lib/ptxd_make_world_inject.sh b/scripts/lib/ptxd_make_world_inject.sh
> > index b74e464c6..90bd8b684 100644
> > --- a/scripts/lib/ptxd_make_world_inject.sh
> > +++ b/scripts/lib/ptxd_make_world_inject.sh
> > @@ -7,10 +7,15 @@
> >  #
> >  
> >  ptxd_make_inject() {
> > -    local source target
> > +    local dest source target
> > +
> > +    dest="${pkg_dir}"
> > +    if [ -n "${pkg_inject_dest}" ]; then
> > +	dest="${pkg_inject_dest}"
> > +    fi
> 
> This should be in ptxd_make_world_inject(). It remains the same for all
> files.

I thought about this when changing it, but decided to keep it at this
place.  Moving it to ptxd_make_world_inject() would require to
introduce another non-local variable and I did not want that.  But
sure, can be done.

> 
> >  
> >      source="$(echo ${inject_file} | cut -d ":" -f 1)"
> > -    target="${pkg_dir}/$(echo ${inject_file} | cut -d ":" -f 2)"
> > +    target="${dest}/$(echo ${inject_file} | cut -d ":" -f 2)"
> >  
> >      if [[ "${source}" =~ ^/.* ]]; then
> >  	ptxd_bailout "'${source}' must not be an absolute path!" \
> > @@ -32,8 +37,17 @@ export -f ptxd_make_inject
> >  ptxd_make_world_inject() {
> >      ptxd_make_world_init || return
> >  
> > -    if [ -z "${pkg_dir}" ]; then
> > -	ptxd_bailout "<PKG>_DIR empty, no destination to inject to."
> > +    if [ -z "${pkg_inject_dest}" ] && [ -z "${pkg_dir}" ]; then
> > +	ptxd_bailout "No destination to inject to." \
> > +	    "Set either <PKG>_DIR or <PKG>_INJECT_DEST to have a valid destination!"
> > +    fi
> > +
> > +    if [ -n "${pkg_inject_dest}" ]; then
> > +	if [ "${pkg_build_dir}" = "${pkg_inject_dest}" ] && [ "${pkg_build_oot}" = "YES" ]; then
> 
> This gives a warning even in the good case, right?

No it does not.  In the good case user has set either <PKG>_DIR or
<PKG>_INJECT_DEST to a not empty string and the bailout does not
happen.

The warning inside this more complicated if statement is just covering
this one edge case which I ran into when developing, and the text
should describe what happens, why the user would not find the files in
<PKG>_BUILD_DIR in that case.

The mkdir is only dependend on <PKG>_INJECT_DEST set, because there's
no need to create that dir if the feature is not used.

> I think we should require that pkg_inject_dest already exists at
> this point and fail if it does not.

This would move the responsibility to create that dir to the package
author.  I opted against that, because <PKG>_DIR and <PKG>_BUILD_DIR
are also implicitly created by ptxdist, no need to create those in the
make rules.  This is a question of policy, but it would make the
inject mechanism somewhat asymmetric if <PKG>_DIR (for when
_INJECT_DEST is not set) has not to be created by the package author
but <PKG>_INJECT_DEST has to.

Again: sure, can be done if desired.

Let me know if I should change the pieces or if my explanations made
sense and things can be kept like this?

Greets
Alex

> Michael
> 
> > +		ptxd_warning "<PKG>_INJECT_DEST is set to <PKG>_BUILD_DIR and <PKG>_BUILD_OOT is set to 'YES'." \
> > +		    "If you called world/inject before world/prepare your files will be removed after injecting."
> > +	fi
> > +	mkdir -p -- "${pkg_inject_dest}"
> >      fi
> >  
> >      for inject_file in ${pkg_inject_files}; do
> > -- 
> > 2.39.2
> > 
> > 
> > 
> 
> -- 
> Pengutronix e.K.                           |                             |
> Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
> 31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
> Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |
>