Zsh Mailing List Archive
Messages sorted by: Reverse Date, Date, Thread, Author

Re: [BUG] SIGSEGV under certain circumstances



On 5 March 2017 at 07:11, Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx> wrote:
> On Mar 1, 11:38pm, Chi-Hsuan Yen wrote:
> }
> } My zsh version is d00931de5c2b7aa846daf137865dd05ac8d4be8a.
> }
> } By the way, after those steps ~/.zsh_history seems corrupted:
>
> I'm not sure the history is actually corrupted, because the history
> file is stored with metafied characters which means you can't easily
> examine it independently of reading it into zsh.  There was a small
> C program posted here some months ago that would unmetafy the file.
>
> However based on your steps to reproduce there likely is a problem
> related to reading the history file back in again.
>
> The stack trace you provided is not helpful because it reveals the
> point at which the corrupted memory is accessed but not the point
> at which it became corrupted.  A pass through valgrind might give
> better information.

(I forgot that gmail servers are blocked, so I re-post my logs)

Thanks for the tip. I re-compiled with commit
8522e996ecc88697344dcc4814367ec7e32e7deb and run zsh under valgrind.
The log can be found in [1]. (The GDB backtrace is also updated to the
new build for consistency). [2] and [3] contains logs of vgdb, which I
heard that can give more information than either valgrind or gdb.

[1] https://gist.github.com/yan12125/014c7a7510d1d9bd9ac8edf142a0c65d#file-valgrind-log
[2] https://gist.github.com/yan12125/014c7a7510d1d9bd9ac8edf142a0c65d#file-valgrind-log-with-vgdb
[3] https://gist.github.com/yan12125/014c7a7510d1d9bd9ac8edf142a0c65d#file-gdb-backtrace-from-vgdb



Messages sorted by: Reverse Date, Date, Thread, Author