Hi,
while working on a new project I discovered that the recent introduction
of "waldux" (?) in the pretty massive commit
9bb871a0bb4c239239944d28bd4d5cfa19d84f62 breaks the kernel download
mechanism for the raspberry pi.
The error seems to be related to double-quotes around the hash. These
confuse the mechanism in fetch.mk that tries to work out a way to
generate the hash-named archive. It will then fail to extract the sources.
You can reproduce by deep-cloning
https://github.com/deets/pififorwarder.git
via
git clone --recursive
then change into the modules/openadk folder and check out the above
mentioned command. Then just run ../../scripts/build.py, and you should
get a non-working build.
I'm also having troubles enforcing GLIBC through the task that you can
find in configs/tasks. Regardless of what I try to put in there (after
manually setting the libc to glibc) - the task can't seem to control
this choice. I noticed a similar problem with the kernel configuration:
the task can't override the selected kernel anymore. It can only be
affected by running menuconfig and manually changing this :( This is not
visible in the repository, but I can reproduce it if you like. Any
suggestions on how to overcome this?
Cheers,
Diez
Hi Martin,
thanks for your answer.
I’m quite aware of git’s capabilities, and we use it extensively in our projects. And in fact we do track changes to OpenADK using git.
However this is still cumbersome when openadk moves on, and our changes risk becoming incompatible. Thus the attempt to grow openadk’s abilities to have less reliance on forking.
I am right now working on a way to run a script after the build has finished. I’ll create a PR and then it can be decided if this is a worthwhile idea or not.
And regarding cfgfs: my question targets build-time, not runtime in individual devices.
Cheers,
Diez
> I am not sure, if I got your problem right. But git provides facilities to only pick what you need from remote. You can create a branch with your changes and pick new stuff from the master.
>
> #einzelne Dateien:
> git checkout <BRANCH> -- <DATAI>
>
> #gesamter Ordner:
> git checkout <BRANCH> -- <ORDNER>/*
>
> You can also pick only some lines:
> git checkout -p Quellenbranch -- path/to/file
>
> Quellenbranch is normally origin/master.
> Changes for a particular system get's committed to cfgfs
>
>
>
> Regards
>
>
> Martin
>
> Am 29.11.2017 um 18:52 schrieb Diez B. Roggisch:
>> Hi all,
>>
>> I’m facing a bit of a conundrum, and would like to discuss this with other users so we can arrive at a hopefully mutually beneficial solution.
>>
>> I’m currently working on getting a raspberry pi compute module 3 to work as a so-called ethernet gadget. This means I have to use quite a range of facilities:
>>
>> - loading specific modules. But not always! Only on demand.
>> - using mdev, with custom rules
>> - using avahi-autopid
>> - a whole bunch of scripts (I prefer python, because I really don’t like bash/shell if I don’t have to use it)
>>
>> Now this touches files mostly in /etc, and some of the problems (like loading the modules deferred) could benefit from inittab or init.d integration. Of course I can (and already do) use the extra-folder to customise all things in /etc. But this means that I lose the ability to profit from newer OpenADK versions or certain features. E.g. a generated rc.conf.
>>
>> So I wonder how best to achieve this? For the sake of the argument lets assume I can’t write my own OpenADK package due to lack of genericness.
>>
>> My current thinking would some kind of hook into either make, or latest the install.sh (not toooo thrilled about that, because it’s huge and I need understanding it better).
>>
>> Any other suggestions, or pointers to something I’m missing?
>>
>> Cheers,
>>
>> Diez
>>
>>
>> _______________________________________________
>> dev mailing list
>> dev(a)openadk.org <mailto:dev@openadk.org>
>> https://mailman.openadk.org/cgi-bin/mailman/listinfo/dev <https://mailman.openadk.org/cgi-bin/mailman/listinfo/dev>
Hi all,
I’m facing a bit of a conundrum, and would like to discuss this with other users so we can arrive at a hopefully mutually beneficial solution.
I’m currently working on getting a raspberry pi compute module 3 to work as a so-called ethernet gadget. This means I have to use quite a range of facilities:
- loading specific modules. But not always! Only on demand.
- using mdev, with custom rules
- using avahi-autopid
- a whole bunch of scripts (I prefer python, because I really don’t like bash/shell if I don’t have to use it)
Now this touches files mostly in /etc, and some of the problems (like loading the modules deferred) could benefit from inittab or init.d integration. Of course I can (and already do) use the extra-folder to customise all things in /etc. But this means that I lose the ability to profit from newer OpenADK versions or certain features. E.g. a generated rc.conf.
So I wonder how best to achieve this? For the sake of the argument lets assume I can’t write my own OpenADK package due to lack of genericness.
My current thinking would some kind of hook into either make, or latest the install.sh (not toooo thrilled about that, because it’s huge and I need understanding it better).
Any other suggestions, or pointers to something I’m missing?
Cheers,
Diez
Hi,
Buildroot and OpenADK have samples to create a Linux system to be
bootup in Qemu system emulation for microblaze architecture.
With gcc 6.3 and 7.1 the samples are not working anymore,
because the Linux system userland does not boot.
Qemu 2.9.0:
Kernel panic - not syncing: Attempted to kill init!
exitcode=0x0000000b
(with glibc, musl and uClibc-ng toolchains)
I bisected gcc source code and found the bad commit:
6dcad60c0ef48af584395a40feeb256fb82986a8
When reverting the change, gcc 6.3 and 7.1 produces working
Linux rootfs again.
What can we do about it?
best regards
Waldemar
