Hello, everyone!
First of all, Apologies for the little delay in response. I was actually a
little stuck with assignments in classes.
Thank you so much, Chris, for such a nice explanation and detailed review.
It cleared a lot of things even for me.
I have tried to respond to further suggestions and feedback below.
On Mon, Aug 24, 2020 at 6:25 AM Chris Johns <chr...@rtems.org> wrote:
> On 23/8/20 8:08 am, Mritunjay Sharma wrote:
> > [Good news and Update]:
> >
> > Thank you so much Chris! Your advice to search for macros using --trace
> solved
> > the problem of hard coding!
> > It took two complete days to figure out this beautiful thing but it is
> every
> > worth it.
>
> Well done, that looks like time worth spending given the results. The
> --trace is
> not the most pleasant of output but it has all the detail.
>
Thanks a lot :)
>
> > Now the user has to just enter the below command and it will make things
> work:
> >
> > `../source-builder/sb-builder --with-rtems-bsp="xilinx_zynq_a9_qemu"
> > --log=log_epics epics-7-1 --trace
> --prefix=$HOME/development/rtems/5-arm`
> >
> > Note: prefix will depend on the user.
> >
> > What made this possible? Your suggestion and the changes in code
> mentioned below:
> >
> > ```diff --git a/rtems/config/epics/epics-7-1.cfg
> b/rtems/config/epics/epics-7-1.cfg
> > index aeb39a9..4b20f82 100644
> > --- a/rtems/config/epics/epics-7-1.cfg
> > +++ b/rtems/config/epics/epics-7-1.cfg
> > @@ -6,6 +6,8 @@
> > %define release 1
> > %endif
> >
> > +%include %{_configdir}/rtems-bsp.cfg
>
> Great.
>
Thanks.
>
> > +
> > #
> > # EPICS Version
> > #
> > diff --git a/source-builder/config/epics-7-1.cfg
> > b/source-builder/config/epics-7-1.cfg
> > index a9581a2..a47aecb 100644
> > --- a/source-builder/config/epics-7-1.cfg
> > +++ b/source-builder/config/epics-7-1.cfg
> > @@ -40,7 +40,10 @@ URL: https://epics.mpg.de/
> >
> > %{build_build_flags}
> >
> > - %{__make} PREFIX=%{_prefix} RTEMS_BASE=$HOME/development/rtems/5-arm
> > RTEMS_VERSION=5 CROSS_COMPILER_TARGET_ARCHS=RTEMS-xilinx_zynq_a9_qemu
> > + #
> > + # Using macros to dynamically path to RTEMS_BASE and RTEMS_VERSION
> using
> > --with-rtems-bsp
> > + #
> > + %{__make} PREFIX=%{_prefix} RTEMS_BASE=%{_exec_prefix}
> > RTEMS_VERSION=%{rtems_version}
> >
> > cd ${build_top}
> >
> > @@ -50,6 +53,5 @@ URL: https://epics.mpg.de/
> > rm -rf $SB_BUILD_ROOT
> >
> > cd ${source_dir_epics}
> > - %{__make} DESTDIR=$SB_BUILD_ROOT PREFIX=%{_prefix}
> > RTEMS_BASE=$HOME/development/rtems/5-arm RTEMS_VERSION=5
> > CROSS_COMPILER_TARGET_ARCHS=RTEMS-xilinx_zynq_a9_qemu
> > -
> > + %{__make} DESTDIR=$SB_BUILD_ROOT PREFIX=%{_prefix}
> RTEMS_BASE=%{_exec_prefix}
> > RTEMS_VERSION=%{rtems_version}
>
> Nice, this is along the lines of what I had hope at the start was possible.
>
> > cd ${build_top}
> > ```
> >
> > These couple of changes made this build successful using RSB recipe. The
> only
> > problem remains
> > is that a warning of 'no hash found' is coming, even though I have added
> it.
>
> The config hash is for epics-%{epics_version}.tar.gz and %source is using
> --rsb-file=epics-base-%{epics_version}.tar.gz. The names do not match. Is
> --rsb-file needed? This is only added if the download file is some
> generated
> name and not what we expect.
>
> > The above changes can be found pushed
> > here:
> https://github.com/RTEMS/rtems-source-builder/commit/a505877157f63f6ae17906276b3ffcb699ed1297
>
> Can you please create a suitable patch against the RSB master repo?
>
Done. I have just sent, please do review.
Modified the patch to remove warning and made a change to ask for rtems-bsp
and host from the terminal.
There's one problem although, while building using:
`./source-builder/sb-set-builder --with-rtems-bsp="xilinx_zynq_a9_qemu"
--log=log_epics epics/epics-base --trace
--prefix=$HOME/development/rtems/5-arm --host=arm-rtems5
`
The EPICS is building perfectly fine but the Buildset fails with the
following error:
`cleaning: epics-base-7.0-arm-rtems5-1
error: copying tree (4):
/home/mritunjay/development/rtems/src/rsb/rtems/build/tmp/epics-base-7.0-arm-rtems5-1-1000
->
/home/mritunjay/development/rtems/src/rsb/rtems/build/tmp/sb-1000/epics/epics-base:
[Errno 2] No such file or directory:
'/home/mritunjay/development/rtems/src/rsb/rtems/build/tmp/epics-base-7.0-arm-rtems5-1-1000'
Build Set: Time 0:02:57.099708
Build FAILED
`
I am not sure what it is and it was not encountered earlier?
> Has your blog been kept updated?
>
Yes, I am submitting my final report blog today which will have the
updates.
>
> > Please do give feedback and suggest improvements so that the recipe can
> become
> > mergeable.
>
> I think it is time to expand the user base and to look at merging the
> script
> into the RSB's master repo.
>
> The next problem to solve is the EPICS's `make runtests`. This is gong to
> be
> hard because it is a meeting of 2 different approaches and I suspect it
> will
> require insight and imagination to find a workable solution.
>
> A test run like `make runtest` is complicated to wrap because the test
> framework
> in EPICS may be needed to support the tests. I do not think you can simply
> copy
> and run a single executable. This is understandable and fine however it
> means
> the way to invoke the tests for a specific RTEMS target needs to be
> embedded in
> to EPICS and this may conflict with the RSB's ability to build EPICS for a
> range
> of BSPs that EPICS has no direct support for. I am not asking you add the
> support to EPICS, rather can we interface to EPICS so we can wrap and
> control
> the pieces we need.
>
> RTEMS has the `rtems-run` command. It is a way to run an executable on a
> target.
> The `rtems-test` command internally wraps the `rtems-run` command to run an
> RTEMS kernel test executable. The RTEMS User manual has more detail.
>
> Is it possible to provide EPICS with a command, make variable or something
> that
> lets us control how an EPICS test executable is executed? If we can join an
> EPICS test run with `rtems-run` we make available the same hardware and
> process
> we use to test the RTEMS kernel to EPICS
>
> I am not sure `rtems-run` as it stands will work as EPICS may be running a
> number of networked executables. I have no idea about this or any insight
> into
> what EPICS is doing here so please have a look and report back.
>
> Another piece of work is to have the RSB script point to EPCIS master
> repo. This
> will mean working with Heinz to get changes in his repo into the upsream
> EPICS repo.
>
>
Have noted all these suggestions and will be starting a thread soon in
relation to them.
Thanks
Mritunjay Sharma
> Chris
>
_______________________________________________
devel mailing list
devel@rtems.org
http://lists.rtems.org/mailman/listinfo/devel
