I
have a comment and a question about this advice, but
probably I am parsing too much as a new user.
I
recently tried to “sync” my development workstations (in my
lab): Gnu Radio 3.5.0rc0 and UHD 3.3.1 which is installed
and working very well on my principal development machine.
A
few days later on another system totally different than my
own, I tried to install new development environments and
initially chose NOT to use Marcus’ build script in full, but
used its “gitfetch” function, followed by:
(cd
to UHD directory)
(git
checkout tag release_003_003_001)
(build-gnuradio
with options to build only the UHD)
And
then,
(cd
to GNU radio directory)
(git
checkout tag 3.5.0rc0)
(build-gnuradio
with options to build only the GNU Radio)
All
of these were on a updated Ubuntu 10.04 LTS. And while the
UHD 3.3.1 system worked initially (uhd_find_device worked
great), it broke immediately after gnu-radio was installed
.. with a segmentation fault as the UHD installed now showed
UHD_003.004.000 instead of UHD_003_003_001. This IMHO is
not what I expected, it should be possible to roll back
(UHD, GnuRadio) to whatever version we wish to have, to test
out older scripts..
Based
upon a suggestion from a colleague, I later ran the
build-gnuradio script in full and all is fine for now,
albeit two different foundation framework.
3.5.0rc0
+ UHD 3.3.1
3.5.0rc0
+ UHD 3.4.0
But,
may I ask the developers to explain the reasoning for this
behavior of an install script that was run on two similar
machines separated by only four days? What can we do,
explicitly, to choose particular <matching>
distributions of Gnuradio and UHD, so we don’t end up having
several versions of the codebase which may have different
classes of bugs. What is your recommended strategy for
regression testing, or synching the various development
environments ?
Regards
s