|
Some OpenOffice.org Hacking Notes
1. Real-time event 32.Program received signal SIG32, Real-time event 32. If gdb says this when you are debugging OOo then its likely that
/lib/libpthread.so is stripped, check with
> file /lib/libpthread*
if the output contains "stripped", then this is your problem. Rebuild glibc,
and this time don't strip the binaries, e.g. for gentoo add nostrip to FEATURES
1. gdb next acts like step because of inline methods You'll see with gdb that when you next it acts like step when inline methods
are reference in the line you want to next over. To resolve this try adding
-O0 instead of -O to the NOOPT options in solenv/inc/unxlngi4.mk.
Additionally, building with -fno-default-inline added to your debugging flags
for linux should resolve it, but this can lead to linking problems.
Nevertheless, to give that a try edit solenv/inc/unxlngi4.mk and add
-fno-default-inline after -g to CFLAGSDEBUG
2. using memprof with OpenOffice.org memprof has problems currently with pthreads, so if openoffice.org hangs with
memprof then
export LD_ASSUME_KERNEL=2.2.5
and try that, currently memprof may have problems catching the application on exit, as
a workaround you can place a
fprintf(stderr, "FOOBAR\n");
sleep(1000000)
at the end of main as found in vcl/unx/source/plugadapt and click memprofs
leak button when you see your FOOBAR printed.
3. making a single file self extracting install set Add a line 'fatbinary name-of-the-file' in instsetoo/util/openoffice.lst
within the Settings section (e.g. below the 'script setup_osl.ins' line)
4. gdb and c++
a) Constructors and breakpoints don't work, see http://sources.redhat.com/cgi-bin/gnatsweb.pl with ids 1091 & 1193 and http://bugzilla.redhat.com/bugzilla/process_bug.cgi b) Deferred breakpoints don't allow deferring on member functions, see http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=128620 c) Deferred breakpoints only work the first cycle of debugging, see http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=128618
|