udev_enumerate_new is called from the latest hidapi code.
Some distros provide a hidapi package. So the actual fcdproplus tries to
detect if an hidapi package is installed. Otherwise it uses the hidapi
code provided with fcdproplus, which uses udev_enumerate_new.
So it would be interesting waht cmake has reported.
Near the end of the cmake output you see something like this ( in the
case that the hidapi package is installed):
GNURADIO_PMT_FOUND = TRUE
-- checking for module 'alsa'
-- found alsa, version 1.0.28
-- Found ALSA 1.0.28
-- checking for module 'libusb-1.0'
-- found libusb-1.0, version 1.0.19
-- Found libusb-1.0: /usr/include/libusb-1.0, /usr/lib64/libusb-1.0.so
-- Found Doxygen: /usr/bin/doxygen (found version "1.8.5")
-- System Hidapi Lib /usr/lib64/libhidapi-libusb.so is used
-- Audio LIBS: /usr/local/gnuradio-orig/lib/libgnuradio-audio.so
As Marcus already proposed, it might be helpfull to see the output of
inside the build directory
and additionally from the gnuradio install directory.
I think udev_enumerate_new is provided from newer versions of libudev.
If you want to use a
hidapi version without udev_enumerate_new
just checkout the with_hidapi release.
Am 01.11.2014 um 01:39 schrieb Marcus Müller:
> Tracing this down; Daniel, could you share what
> pkg-config --static --libs libsub-1.0
> in your build directory, can you do a
> ldd lib/libgnuradio-fcdproplus.so
> Greetings, and good night (forgot the time)
> On 11/01/2014 01:04 AM, Marcus Müller wrote:
>> Oh indeed, good point!
>> Sylvain: you're right.
>> How gr-fcdproplus ends up being able to compile and link is still a
>> mystery to me. I was expecting a horrible "symbol export forwarding
>> hack" involving libusb, but no:
>> $>nm -D /lib64/libusb-1.so
>> U udev_device_get_action
>> U udev_device_get_devnode
>> U udev_device_get_sysname
>> U udev_device_new_from_syspath
>> U udev_device_unref
>> U udev_enumerate_add_match_subsystem
>> U udev_enumerate_get_list_entry
>> U udev_enumerate_new
>> gives me all of the udev symbols as undefined, which obviously is
>> correct, because they should be dynamically loaded from libusb:
>> $>ldd /lib64/libusb-1.0.so
>> libusb-1.0.so|grep udev
>> linux-vdso.so.1 => (0x00007fff6b7fc000)
>> libudev.so.1 (0x00000038b6600000)
>> Trying to build this now, but I'm currently on a machine where I
>> haven't even installed GR yet :/ this might take a second.
>> On 11/01/2014 12:43 AM, Sylvain Munaut wrote:
>>> What I find strange is that the symbol error is found at runtime and
>>> not at link-time. That suggest either there is two libudev and the one
>>> used for runtime is different than the one found for link-time, or
>>> that this library was copied over from another system.
>> Discuss-gnuradio mailing list