Thank you for the response, Nicolas. I realized the aec2fe9 commit from
27-JUN-17 only applies to the "maint" branch in the "fpga" repo. Instead I
should checkout commits from the "rfnoc-devel" branch of the "fpga,"
"gr-ettus," and "uhd" repos and "master" branch of the "gnuradio" repo
(from gnuradio/gnuradio github, not ettusresearch/gnuradio). So after
uninstalling whatever builds I had of uhd, gnuradio, and gr-ettus (I guess
uhd-fpga isn't installed), I checked out commits on or before 27-JUN-17:
cd uhd
git checkout ba11bf2d699957e9a5d9b587de045162d2fce3e3
cd gnuradio
git checkout 11d59e9f227e3d3f5c388c376ee59e0cfa021dcf
git submodule update --init --recursive
cd gr-ettus
git checkout fbea7f763460ffb42592c6c1c389047b6e0e1a10
cd uhd-fpga
git checkout 8b731c3b277fb2caee0b9633c9b159e55f175486
Then built in that order but I get a make error in "uhd":
/home/switchlanez/rfnoc/src/uhd/host/include/uhd/rfnoc/block_ctrl_base.hpp:74:69:
error: invalid new-expression of abstract class type
‘uhd::rfnoc::magnesium_radio_ctrl_impl’
return block_ctrl_base::sptr(new CLASS_NAME##_impl(make_args)); \
...
lib/CMakeFiles/uhd.dir/build.make:3088: recipe for target
'lib/CMakeFiles/uhd.dir/usrp/dboard/magnesium/magnesium_radio_ctrl_impl.cpp.o'
failed
I guess I'll try a different commit from "uhd" repo/"rfnoc-devel" branch
unless you or someone can see something else I'm doing fundamentally wrong.
Andrew
On Sun, Apr 29, 2018 at 10:40 AM, Nicolas Cuervo <[email protected]>
wrote:
> Hello Andrew,
>
> I did a fresh install of the latest rfnoc-devel (using pybombs prefix init
>> ~/rfnoc -R rfnoc -a rfnoc) and RFNoC blocks do not work (such as FIR and Rx
>> Radio, possibly related to USRP-users mailer subject: Core dump with
>> UHD_3.11, X310, and LFTX). So I am trying to checkout and build rfnoc-devel
>> to a known stable commit: aec2fe9. I ran git checkout aec2fe9 in
>> rfnoc/src/uhd, rfnoc/src/gnuradio, and rfnoc/src/gr-ettus. Then from
>> uhd/host/build I ran cmake ../ but got this error:
>>
>>
>>
>> CMake Error: The source directory "/home/switchlanez/rfnoc/src/uhd/host"
>> does not appear to contain CMakeLists.txt
>>
>
> Please correct me if I'm wrong, but did you run "git checkout aec2fe9" in
> *all* those repositories? that Git hash might be correct for only one of
> them, and the corresponding according commit in the other repositories
> remains unknown. Where did you take that hash from?
>
> If what I understood is what you did, that might be causing you problems.
> You need to identify which commits for each of the repositories is
> according to an "stable" version, and checkout accordinglu and respectively.
>
>
>> So I downloaded that file from GitHub with the closest date on or before
>> aec2fe9 was released and tried cmake ../ again. But then it could not find
>> a bunch of load files: UHDComponent, UHDPackage, etc. And the cycle
>> continues.
>>
>
> See my previous comment.
>
>
>>
>> 1) What am I doing wrong here? Reverting to older commits seems like a
>> typical thing RNFoC users would do but I haven't found specific steps in
>> the mailer (links to topics would help).
>>
>
> See my previous comment. Reverting to older commits is something that is
> commonly done, yes. From what I understood you run to a same hash in all of
> the involved repos. I might have understood wrong, but that's not what
> you'd want to do.
>
>>
>>
>> 2) I was told the maint branch head will be updated "very soon." Assuming
>> I do another fresh new install (using pybombs prefix init ~/rfnoc -R rfnoc
>> -a rfnoc like I did above) but with a different prefix, how would the steps
>> be different to start from the latest rfnoc-devel installation and update
>> it to the maint branch head? Or is there a pybombs recipe to build the
>> latest rfnoc-maint?
>>
>
> There is no "rfnoc-maint". All RFNoC development is done in the
> "rfnoc-devel" branch. If you change the branch, you wouldn't be able to use
> all the RFNoC features needed for development.
>
>
> Cheers,
> - Nicolas
>
_______________________________________________
USRP-users mailing list
[email protected]
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
