vi
commented
10 years ago Seems like it still happens with closed Navigator as well...
Closed vi closed 10 years ago
vi
commented
10 years ago Seems like it still happens with closed Navigator as well...
vi
commented
10 years ago MALLOC_CHECK_=1 does not help.
andreikop
commented
10 years ago Hi
Could you test it on Enki master + Qutepart master?
vi
commented
10 years ago Updated:
Version 14.03.0
Uses Qutepart 2.1.0 (with binary parser)Still crashes, the same backtrace.
andreikop
commented
10 years ago cParser.so and test without it? (with binary parser should disappear and performance will be worse)vi
commented
10 years ago Reproduced without cParser.so (Uses Qutepart 2.1.0 (with Python parser)), the same backtrace. There are also additional repeated ERROR:qutepart:#pop value is too big 3 unlike with native version.
Shall I try to get debugging symbols for Python and PyQt to make nicer backtrace?
vi
commented
10 years ago I tried installing debugging packages for python and python-qt4; this caused both packages to be upgraded; this in turn caused the bug to stop being reproducible.
Probably the bug was not Enki-specific.
vi
commented
10 years ago Reproduced again, now with debugging packages installed.
Mega-backtrace: https://gist.github.com/vi/cc46cf95cd5b44f7964b
The problem is reproducible even on this small cpp file:
void foo() {
if (qqq) return;
}It opens as Enki starts (along of many other files, some of which are not writable), I try typing logger instead lf qqq, and it crashes (not from the first attempt maybe).
If I try the same with Enki started from other user account (with other session) displaying window on the same X server, it does not crash.
vi
commented
10 years ago Another problem while editing C++ code:
Traceback (most recent call last):
File "/home/vi/src/git/qutepart/qutepart/completer.py", line 469, in _onCompletionListItemSelected
selectedWord = model.words[index]
IndexError: list index out of rangeThe crash: Thank you for the useful backtrace. I haven't reproduced the bug, but committed a bunch of patches, which might fix it. Try to reproduce the bug on Qutepart master.
The exception: I can workaround the bug by adding list range check, but I'd like to understand the problem. Do you have an idea how to reproduce it?
vi
commented
10 years ago I don't know. I was editing the C++ code, the bug triggered instead of showing of code completion window. Maybe the file was incomplete or with syntax errors...
Try to reproduce the bug on Qutepart master.
I'll update it and retry when I see the bug the next time. Now it does not happen.
vi
commented
10 years ago Reproduced again (with older Qutepart). After updating Qutepart it does not crash.
git-bisect points to the fixing commit: hlamer/qutepart@6682f2b5cc5a512fdfca1c782605ed8c1ce218e3 "do not monkeypatch..."
andreikop
commented
10 years ago Cool! Thank you for finding the bug. I'll release v2.1.1
Version:
The problem is reproducible now, but the backtrace is not necessary exact (now I see the address
#0 0x20202020 in ?? ()).The problem appears when I type the second character in new identifier, like this:
(typing "O" crashes Enki).
If I close the Navigator, Enki behaves normally.
One more observation: messages like
may to affect the crash. The crash may happen when the second of two such messages expires (affecting the window layout a bit).
If I follow other sequence of UI actions (other delays between typed text, closing or not closing bottom messages manually, typing other text in other place) it may fail to reproduce or reproduce later...