RTEMS examples build failed

[Sebastian Huber]

On 20/04/2020 19:58, Joel Sherrill wrote:

>
>
> On Mon, Apr 20, 2020 at 12:20 PM Sebastian Huber 
> <sebastian.huber at embedded-brains.de 
> <mailto:sebastian.huber at embedded-brains.de>> wrote:
>
>     On 20/04/2020 19:04, Vijay Kumar Banerjee wrote:
>
>     > On Sun, Apr 19, 2020 at 8:16 PM Sebastian Huber
>     > <sebastian.huber at embedded-brains.de
>     <mailto:sebastian.huber at embedded-brains.de>
>     > <mailto:sebastian.huber at embedded-brains.de
>     <mailto:sebastian.huber at embedded-brains.de>>> wrote:
>     >
>     >     On 19/04/2020 15:38, Gedare Bloom wrote:
>     >
>     >>     On Sun, Apr 19, 2020 at 3:52 AM Vijay Kumar
>     Banerjee<vijay at rtems.org <mailto:vijay at rtems.org>> 
>     <mailto:vijay at rtems.org <mailto:vijay at rtems.org>>  wrote:
>     >>>     Hi,
>     >>>
>     >>>     While trying to build the examples with a fresh pull from
>     upstream,
>     >>>     I got the following build error from gccdeps:
>     >>>
>     >>>     ```
>     >>>     Build failed
>     >>>     Traceback (most recent call last):
>     >>>        File
>     "//home/lunatic/development/rtems-examples//.waf-2.0.19-1f3c580272b15a03d2566843c5fe872a/waflib/Task.py",
>     line 190, in process
>     >>>          self.post_run()
>     >>>        File
>     "/home/lunatic/development/rtems-examples/rtems_waf/gccdeps.py",
>     line 144, in post_run
>     >>>          raise ValueError('could not find %r for %r' % (x, self))
>     >>>     ValueError: could not find ['filesystem', 'fat_ramdisk',
>     'fs-root-tar.h'] for
>     >>     It's doing an error check to find the build dependency
>     fs-root-tar.h,
>     >>     which gets generated in the build directory by bin2c. This
>     has always
>     >>     been tricky to handle in waf. I don't grok the way Chris
>     reworked the
>     >>     rootfs support in the example so I don't immediately see
>     the root
>     >>     (hah) problem or how to try to fix it.
>     >
>     >
>     > Hi,
>     >
>     > I figured out that the tar is generating properly but the error
>     check
>     > is failing
>     > because it's looking for the fs-root-tar.h file in the source file
>     > directory but the
>     > header file is being generated in the build directory.
>     >
>     > Just manually copying the header file to the source directory
>     makes it
>     > happy
>     > and the build works fine. Do we want to generate the header file in
>     > the source
>     > directory as well and maybe add it in gitignore?
>     >
>     > The other approach can be to tell gccdeps to check for the files in
>     > the build
>     > directory instead. I did it with a workaround like the
>     following, and
>     > it works fine:
>     > ```
>     > diff --git a/gccdeps.py b/gccdeps.py
>     > index bfabe72..61d9bfd 100644
>     > --- a/gccdeps.py
>     > +++ b/gccdeps.py
>     > @@ -141,7 +141,10 @@ def post_run(self):
>     >                         node = path_to_node(path, x, cached_nodes)
>     >
>     >                 if not node:
>     > -                       raise ValueError('could not find %r for
>     %r' %
>     > (x, self))
>     > +                       x =
>     > os.path.relpath(str(bld.bldnode)).split('/') + x
>     > +                       node = path_to_node(path, x, cached_nodes)
>     > +                       if not node:
>     > +                           raise ValueError('could not find %r for
>     > %r' % (x, self))
>     >                 if id(node) == id(self.inputs[0]):
>     >                         # ignore the source file, it is already
>     in the
>     > dependencies
>     >                         # this way, successful config tests may be
>     > retrieved from the cache
>     > ```
>     >
>     > If it doesn't look too hackish, then I'll post a patch. This
>     shouldn't
>     > break anything else
>     > and I'll surely add comments there about the use case of taking
>     this
>     > approach.
>     >
>     > Does it look right?
>     I encountered the same error while writing the new build system. I
>     think
>     this issue can be fixed with a proper dependency chain in waf.
>     Patching
>     the generic gccdeps.py is not the way to go if you ask me.
>
>
> You could alternatively copy the rootfs base files to the build 
> directory and
> do the tar from there.  Source directory stays clean.
We don't have to reinvent the wheel. We could copy the wscript of the 
new build system to the examples and add a couple of specification 
items. Building test programs is really not that much different to 
building applications.