#637 Crash after clicking finish on the wxWidgets project creation wizard

[greenbreen]

I'm using codeblocks SVN r11305 configured with "./configure --with-contrib-plugins=wxsmith" on Ubuntu 17.10. If I perform the following steps, codeblocks crashes:

1. Click "Create a new project"
2. Select "wxWidgets project" and click "Go"
3. Leave selection at default "wxWidgets 3.0.x" and click "Next >"
4. Type a name in "Project title:" and click "Next >"
5. Click "Next >" 5 more times to accept the default settings
6. Click "Finish"

A debug report popup is very briefly displayed, but then it and codeblocks closes.

I had gotten the same behavior with the codeblocks 17.12 built from source with the same configuration.

I tried to follow the guidance at https://wiki.ubuntu.com/Backtrace, but I'm not sure I did it right; after the crash gdb is still running but will not accept any commands and does not respond to ctrl+c, so I had to kill it. I've attached the output that I got from it. I've also have the xml debug report that codeblocks created after the crash; I can attach that also if needed.

[Teodor Petrov]

Do you have another version of codeblocks installed on your system?
It seems that the lib is found in /usr/lib/..., but the default prefix in configure is /usr/local, so it should have mentioned that the lib is in /usr/local/lib.

After starting gdb and everything loads can you press ctrl-c and then run "info sharedlibrary" and paste the output?

On ubuntu it is more reliable if you build packages and not use make install. Another good idea is to use non-default prefix (--prefix=/home/myname/software/cb for example).

[greenbreen]

I did have the packaged version of codeblocks installed, but I had problems with it and uninstalled it.

I tried doing the info sharedlibrary thing, but after pressing ctrl-c, gdb does not echo anything I type, and typing "info sharedlibrary" and pressing enter seems to have no effect, although typing "quit" and pressing enter did quit gdb (most of the time). I'm not familiar with gdb, so maybe I'm doing something stupid.

I'll have some time after work to try building codeblocks on a fresh install of ubuntu 17.10 to see if the problem goes away, and I'll report the result here.

[Teodor Petrov]

Executing lsof -p pidof codeblocks could provide similar information...

[greenbreen]

I've attached the output from running lsof. I'm currently backing up my system in preparation for reinstalling Ubuntu, so it may be a while before I can report on whether or not I have the problem with a fresh install.

[Teodor Petrov]

You have this line: /usr/lib/libwxsmithlib.so.0.0.1 in the log.
I think it should be something like: /usr/local/lib/libwxsmithlib.so.0.0.1
Can you verify that this file is from a wrong version of CB?

[greenbreen]

I checked /usr/lib and /usr/local/lib, and both contained libwxsmith.so.0 and libwxsmith.so.0.0.1. I deleted those files from /usr/lib. When I started codeblocks, I got a message about a plugin failing to load, as shown in the attached image. When I run lsof and grep for "smith", nothing comes up.

[Teodor Petrov]

Interesting. I guess something in your library load paths is not correct. Have you run ld-config after you've installed codeblocks with make install?

[greenbreen]

I believe I ran ld-config after I installed with make install the first time, but I might not have done it the second time. The first time I installed codeblocks 17.12 from source that I downloaded as a compressed file off the website. After running into this problem, I ran "sudo make uninstall". I afterwards pulled the source from svn and built it.

I'm currently working through the build process on a fresh install of ubuntu 17.10. It's taking a while to get all the dependencies, etc., straight, but I'll provide an update once I've gotten codeblocks to build successfully.

[greenbreen]

I've been trying without much success to build codeblocks from source. I ended up just installing the packaged version, and it seems to be working ok so far. Since I'm unable to reproduce my original problem, and since it seems to have been more of a configuration problem than a bug anyway, this ticket can probably be closed as invalid unless other people are seeing the same problem. I don't think that's likely, though, because I searched around for similar problems before I opened the ticket, but didn't find any.