Segmentation fault when editing profiles #101
Comments
|
Oh, just another seemingly random crash while terminix was open in the background. Probably something different: Any strategy how I can find the root for this? I just see no pattern to these crashes. |
|
Ok, found out how to get a backtrace via gdb: Unfortunately this is somewhere inside the Gtk event loop and from this output I cannot see which event in terminix caused it. This is using GTK 3.18 on Arch Linux under Gnome. |
|
Thanks for your effort in getting a stack trace for this. I'll look into it
|
|
I just started my work day so I won't be able to look at this for awhile and I suspect I might have some trouble reproducing it, if you have some time can you try running it in gdb using the following line and then reproduce the crash: I'm suspecting a similar issue to what I experienced last time where D is freeing some memory in advance of GLIb. The debug-blocks will hopefully provide more info. |
|
Will do, but I think I won't find the time until tomorrow. The trouble is I still have no clear way to reproduce the crash, I just know it will eventually crash when I am inside the profile editing. I am also not entirely sure if we are dealing with one or maybe multiple sources for the crash. |
|
@phw I fooled around with this at lunch and was able to reproduce it once, the stack trace I get shows a very similar to pattern to a previous memory issue that was resolved in GtkD. It's possible that this resolution was not perfect and requires additional work or I'm doing something incorrect in my code. One thing to understand about GtkD is that most of the code in it is auto-generated from GIR files. As a result, while we may see it happening in different spots it could actually all be the result of one bug where a specific code pattern is repeated across different GtkD classes and structs. I have opened an issue with GtkD for this here: gtkd-developers/GtkD#146 If you do have some time later, I would appreciate it if you could try to reproduce it using the command I posted earlier. The more data points we have the easier it is for myself and the GtkD maintainer to understand the issue. I'll be spending some time myself tonite trying to reproduce it. |
|
Here the gdb output with What I actually did this time was go righ into settings -> profile, default profile edit, close, edit, close, edit. While clicking the last edit it crashed. That is reproducible, but only when run with gdb. Without gdb the same procedure runs just fine, but it will crash eventually nonetheless. |
|
@phw As you probably saw the GtkD maintainer provided a modified version of RGBA.d to workaround the issue. If you are interested in testing it and need me to produce a build for you just let me know. |
|
I won't be able to test this properly the next 4 days, but if you could provide a build or release next week I will test it thoroughly. |
|
No worries, I pushed out a new release so you can grab that when you get a chance and test it. |
|
Closing as this should be fully addressed in the latest binary release, if not please re-open this bug. |
This is a strange one. When editing profiles sometimes terminix crashes with a segmentation fault. I have not yet found out how to reliable reproduce this, but it happens while inside a profile. I tried to reproduce it with the debug build and I got a crash with this output:
Note that I tried repeatedly to edit profiles and do changes until at some point the crash happened. In this particular case the crash happened at the moment I activated the "Custom font" checkbox, but this is not reproducable and in previous crashes I did not use that checkbox at all. Must be something unrelated.
The text was updated successfully, but these errors were encountered: