Re: CVS Compile error, not sure if comedi is happy after a workaround

I made a couple modifications to my installation process, and having 
both rtai and comedi working, now! First, thinking that maybe it wasn't 
so great of an idea to use a kernel that rtai didn't include an adeos 
patch for, I upgraded to 2.4.27, built, compiled, and installed. Second, 
I used rtai 3.1 instead of 3.1r2, because I know that this has worked 
with comedi for at least two other labs. After installing that with 
comedi-0.7.70 and comedilib-0.7.22, everything runs brilliantly. I was 
still having the same problem with CVS Comedilib throwing compile errors 
related to the /doc directory, though, which is why I went back to the 
release. I wonder about this.

My working rtai/comedi combination:
UP AMD Athlon Thunderbird
NI PCI-6052E on a PCI-33MHz bus
Mandrake 9.2
vanilla 2.4.27 kernel patched with hal patch included with rtai-3.1
comedi 0.7.70
comedilib 0.7.22

Now if I could just figure out why comedi won't work on my SMP Opteron 
box...

Regards,
Ryan

Ryan Hooper wrote:

> Hi all,
> It seems that my opteron machine doesn't want to run comedi because of 
> a     hardware issue, so I've been trying out comedi on an older 
> machine of mine (a single 1.1GHz Athlon Thunderbird, PCI-33MHz ports, 
> mandrake 9.2, gcc 3.3.1, etc.). I installed comedi 0.7.70 on an 
> unmodified vanilla 2.4.22 kernel yesterday, and the whole thing seemed 
> to work great. No compile errors, dmesg looked good, and comedi_test 
> ran great.
>
> Next I wanted to get comedi working with rtai, so today I installed 
> rtai 3.1r2 on an adeos r16 patched vanilla 2.4.22 kernel, and rtai 
> seems to work great. Then I tried to install comedi, hoping it would 
> work with rtai, and that's when my troubles started. After sucessfully 
> configuring comedi, I ran make, which aborted with the following build 
> log:
>
> ......cut.........
> -------------------------------------------------------------------
>
> I then downloaded, configured (after upgrading to automake 1.7.6 and 
> autoconf 2.57), and ran make on CVS, but obtained the same result.
>
> To try to get things going, I commented out "typedef void irqreturn_t" 
> in $rtai-installdir/include/rtai_wrappers.h, which allowed comedi to 
> build and install (after which I built and installed comedilib CVS, 
> which threw a bunch of compile errors when trying to build everything 
> in the /docs directory, which I addressed by removing docs from the 
> Makefile SUBDIRS), but I'm not sure if the resulting installation is 
> behaving right. For one thing, the results of comedi_test throw a good 
> number more errors than the results for the NI PCI-6052 as listed in 
> CVS comedilib/testing/results. Also, I get this dmesg after I run 
> comedi_config /dev/comedi0 ni_pcimio:
>
.....cut.....

>
> Does anyone have any thoughts about what's going on with either the 
> compile error or that last line of dmesg? I'm guessing any problem 
> caused in the installation stems from the removal of the typedef of 
> irqreturn_t in rtai_wrappers.h, but I have no idea of what this 
> variable does, beyond that, or where it is previously (mistakenly?) 
> defined, if that's the real reason it is throwing a compile error.
>
>
> Regards,
> Ryan Hooper
> Emory University Biology Dpt.
> ryan.hooper_at_emory.edu
>
> _______________________________________________
> comedi mailing list
> comedi_at_comedi.org
> https://cvs.comedi.org/cgi-bin/mailman/listinfo/comedi

Received on 2005-08-03Z21:00:48