cscope crashes

[hnanni]

cscope is crashing from the terminal:

[607249.470513] hnanni-desk3.*** systemd-coredump[656763]: Process 614535 (cscope) of user 1000 dumped core.

                                                                         Stack trace of thread 614535:
                                                                         #0  0x00007f163c54b6f1 werase (libncurses.so.6 + 0x116f1)
                                                                         #1  0x00005652f730ca55 display (cscope + 0x11a55)
                                                                         #2  0x00005652f72ffc3c main (cscope + 0x4c3c)
                                                                         #3  0x00007f163c33351d __libc_start_main (libc.so.6 + 0x2a51d)
                                                                         #4  0x00005652f73007fe _start (cscope + 0x57fe)
-- Subject: Il processo 614535 (cscope) ha generato un dump.

The cscope issue is not observed if ncurses is rebuilt not using termlib on its config i.e.

diff --git a/configure b/configure
index dc312a4..1753443 100644
--- a/configure
+++ b/configure
@@ -1 +1 @@
---with-shared --with-termlib --enable-widec --without-progs
+--with-shared --enable-widec --without-progs

[hnanni]

BTW, ncurses is on version 6.2. Maybe an update is due in CLR (6.2 it crashes the same as current ncurses when --with-termlib is in the configure params )

[ahkok]

I don't see cscope crashing - any particular action you're doing?

I'm not sure we want to disable termlib, either. Updating ncurses maybe, of course.

[hnanni]

I'm connecting trough ssh to the machine that executes cscope and stores the code. Issuing "cscope -q -p 4 -R -k" from both, gnome-terminal or urxvt and it crashes as described. When adding -b, the indexing is completed without problem.

[hnanni]

It seems ssh is not playing. Got the code in the local machine and got the cscope crash. The crash is gone again with the libncurses built wo the terminfo

[ahkok]

I would thing though that a coredump in cscope is more likely a bug in cscope (incorrect use of API, e.g.) than a bug in ncurses.

[hnanni]

yes. The ncurses observation is just a piece of data and the workaround I'm using to manage the issue

[hnanni]

where are the core files landing? How are the debug symbols acquired?

[thiagomacieira]

Core files are captured by systemd-coredumpd. Use coredumpctl to list them, export them or start gdb on them.

Debug symbols are automatically downloaded by the clr_debug_fuse service.