[Asterisk-Dev] whee, floating point exception with app_rxfax

[Andrew Kohlsmith]

On January 13, 2005 12:10 pm, Steve Underwood wrote:
> It has failed in process_baud, but that is a fairly big routine doing a
> lot of floating point. V.27ter is the slow modem, used for the 2400 and
> 4800bps FAX modes. That is not used that much, so it has received less
> user testing than the V.29 code. It has been tested just as thoroughly
> here, though, and I never had a failure. Is this failure repeatable? If
> so, you need to build the code with the -g option to GCC, so it will
> tell you the actual source code line at which the error occured.

just an update:

libspandsp.so.0.0.0 appears to already be compiled with -g, as the Makefile 
seems to use CFLAGS=-g -O2, and as I mentioned before the 
'file /usr/lib/libspandsp.so.0.0.0' output says it's not stripped.

However running "nm -al /usr/lib/libspandsp.so.0.0.0 | grep ' N '" shows only 
four symbols without names, just as gdb shows it.

In fact, make clean and CFLAGS=-g make changes nothing (I see -g used 
throughout the build process output), and just to make sure it wasn't getting 
lost in the link stage I ran this:

gcc -shared 
-g .libs/adsi.o .libs/awgn.o .libs/bert.o .libs/complex_dds.o .libs/complex_filters.o .libs/dds.o .libs/ec_disable_detector.o .libs/echo.o .libs/fsk.o .libs/hdlc.o .libs/ima_adpcm.o .libs/oki_adpcm.o .libs/oss.o .libs/power_meter.o .libs/queue.o .libs/sig_tone.o .libs/super_tone_rx.o .libs/super_tone_tx.o .libs/t4.o .libs/t30.o .libs/t31.o .libs/t35.o .libs/testcpuid.o .libs/time_scale.o .libs/tone_detect.o .libs/tone_generate.o .libs/v17rx.o .libs/v17tx.o .libs/v22bis_rx.o .libs/v22bis_tx.o .libs/v27ter_rx.o .libs/v27ter_tx.o .libs/v29rx.o .libs/v29tx.o .libs/v8.o  
-lpthread /usr/lib/libxml2.so -ltiff  -Wl,-soname -Wl,libspandsp.so.0 
-o .libs/libspandsp.so.0.0.0

Nothing new -- four unnamed debug symbols and that's all.

-A.