Hi, I am having a strange issue calling snd_pcm_* alsa-lib functions from a .so file.
The example is a pretty run of the mill .wav file output example in C from the internet, which I have placed in playwav.c. I put the routine that does the pcm play, alsaplaywave(), into a routine and have a duplicate in another external file, test2.c. I named the one in playwav.c alsaplaywav1(), and the one in test2.c alsaplaywav().
The compile is done as follows:
gcc -g3 -Iinclude -fPIC -c test2.c -o test2.o gcc -shared test2.o -o test2.so # # Compiling from .so causes bug, compiling from .o does not # gcc -g3 -Iinclude linux/playwav.c ./test2.so -lasound -o playwav #gcc -g3 -Iinclude linux/playwav.c ./test2.o -lasound -o playwav If the test2.so file is used, the playback has its parameters such as rate messed up, and the result is it plays as fast as possible (series of chirps). If the above line is just changed to test2.o, the playback is fine. Further, if the copy of alsaplaywav() local to the playwav.c file is executed, the playback is correct. It is only when playing from the routine in the .so file that it messes up. To find out what the key point of failure is, I put all of the preamble for the routine up until it does a
snd_pcm_hw_params_set_rate_near() at the calling point (in playwav.c) and commented that out in the routine. The idea was to see exactly which alsa call being done inside the .so caused the malfunction. The rate set appears to be the key. Do that outside the .so, it works, do it inside, it does not. I note that snd_pcm_hw_params_set_rate_near() goes through a few macros in alsa-lib. I think I could debug this better if I can get alsa symbols into gdb. As in the other message, I cannot figure out where the libasound.so gets constructed in the build. I see libasound.a, but no libasound.so. Apologies for not cutting the example down further, but the failure mode is quite complex and that is difficult. Find playwav.c and test2.c attached, and the compile instructions I used above or in cplaywav. The files are fairly small. Thanks for any help.
Scott Franco