Development for DS107+ (and other Marvell ARM based systems)

[yinka]

Well, I use the qemu-arm-0.8.2-sb2 (just because it looks like the most recent )

You do not need to move your libraries around - that's a recipe for disaster! Just add /usr/local/lib to your /etc/ld.so.conf and run ldconfig.

I guess following the tutorial is tedious. Just logging into scratchbox and running the sb-menu is quite straightforward.

HINT: I use firefox so I installed the FireFTP extension. This makes transferring the compiled programs to Diskstation a lot easier for me.

I choose the qemu-arm-0.8.0-m2 . Not a clue if it is right or not

But the bottom line is IT WORKS!!

had to move the libs from /usr/local/lib to /lib but after that Midnight commander is working.

Moving on to rar now

Must say it's a bit tedious as im a total beginner on linux but i'm also quite proud having compiled my first linux file

//Mike

[yinka]

For those interested, the GPL sources are available at the french forum here: http://www.nas-forum.com/wiki/index.php/Sourcesgpl

New GPL sources are now available!

[ed_x]

Wow! Scratchbox is really cool, it even comes with an emulator! I'm trying it right now.

[polarbear]

Is there by any chance a Windows hosted development environment, maybe based on cygwin?

[flipflip]

I use coLinux (and Cygwin). It works perfectly.

[ed_x]

Float don't seem to work properly. Did anyone manage to get floats working? When I try some float test program, it crashes on the DS with an 'Illegal Instruction'. Can't really find how to do a softfloat either...

[ikeke]

Can report that I followed yinkas information and the basic tutorial on Scratchbox. Althoug I had to go into the First chapter as well and run sb-menu to find the right q-emu as the string from the tutorial didn't work. I choose the qemu-arm-0.8.0-m2 . Not a clue if it is right or not

But the bottom line is IT WORKS!!//Mike

Thanks for your help Yinkas and Mike, it works for me too.
Just one question:

I compiled glib, ncurse and nano by using ./configure then make and make install but how do i proceed to put the result in the syno ? is there a way to have a list telling where all the files have been created and copied in the scratchbox environment ?

[m.aicardi]

After a first "wow" having run the "hello world" on my CS407, I tried compiling some simple programs: prozilla and nano.

They correctly run under Scratchbox, but when transferred to the CubeStation this is the result:

Code: Select all

CubeStation> ./nano
Illegal instruction (core dumped)

This is the file information:

Code: Select all

[sbox-CS407: ~/nano/nano/src] > file nano
nano: ELF 32-bit LSB executable, ARM, version 1 (ARM), for GNU/Linux 2.4.3, dynamically linked (uses shared libs), not stripped

Anybody have some clue?

Thanks.

Marco

[Mike]

ikeke;

The suggestion which I got from Ed in this matter as I'm just as large a newbie as you are is to when running ./configure add -prefix=/opt/<package_name>

Then you just tar the binaries which are put there after make and make install. Move them to the Cubestation and untar them in /opt/local (suggestion). Please ensure that your path is updated to also include /opt/local/bin

Good Luck

[ikeke]

ikeke;

The suggestion which I got from Ed in this matter as I'm just as large a newbie as you are is to when running ./configure add -prefix=/opt/<package_name>

Then you just tar the binaries which are put there after make and make install. Move them to the Cubestation and untar them in /opt/local (suggestion). Please ensure that your path is updated to also include /opt/local/bin

Good Luck

Thanks a lot Mike, i'll try this solution
When you say i have to modify the path, i guess that you mean in the Syno ? if so, ok i'll do it.

I find the source of ipkg, and compiled it. it works on scratchbox so if i can install it on the syno, i'll make ipkg package of each apps i compiled

[ed_x]

that's not it. I suspect that scratchbox tries to do hard-float for arm-linux, which won't work on arm-v5tejl

[ldesnogu]

that's not it. I suspect that scratchbox tries to do hard-float for arm-linux, which won't work on arm-v5tejl

My understanding is that Feroceon has hardware FP, similar to ARM standard VFP, but not 100% compatible (look at traffic on Linux ARM kernel mailing-list regarding Orion platform).

[Mike]

Wouldn't it be possible to create a native compiler from source which is compiled on the CS itself to avoid the float problem?

[ldesnogu]

Wouldn't it be possible to create a native compiler from source which is compiled on the CS itself to avoid the float problem?

That would not change the problem *at all*, re-think about what you wrote

As I already wrote somewhere, Marvell and hence Synology use a cross-compiler provided by CodeSourcery.

The undefined instruction might be caused by two problems:
1. our compiler generates an undefined instruction (either the code generator is not OK for Feroceon or we use wrong flags)
2. the C library (or part of the startup code) contains an undefined instruction.

[yinka]

Well, ldesnogu, I do not know much about the technicalities of the VFP code generation issue, but when I did a canadian cross-compiler (by compiling crosstool-4.3 twice) to run natively on the DS107+, the "llegal instruction" problem largely disappeared. Probably some issues with scratchbox also

The ELDK native tools also compile very well on the DS107+, but the libraries are a different version

Wouldn't it be possible to create a native compiler from source which is compiled on the CS itself to avoid the float problem?

That would not change the problem *at all*, re-think about what you wrote

As I already wrote somewhere, Marvell and hence Synology use a cross-compiler provided by CodeSourcery.

The undefined instruction might be caused by two problems:
1. our compiler generates an undefined instruction (either the code generator is not OK for Feroceon or we use wrong flags)
2. the C library (or part of the startup code) contains an undefined instruction.