| commit | 8412796ac1d53570e94748ab96f32d0919908550 | |
| author | Steffen Nurpmeso <steffen@sdaoden.eu> | |
| Thu, 2 Aug 2018 21:37:49 +0000 (2 23:37 +0200) | ||
| committer | Steffen Nurpmeso <steffen@sdaoden.eu> | |
| Thu, 2 Aug 2018 21:58:21 +0000 (2 23:58 +0200) | ||
| tree | a2ecd41065749b71f56f5629544a9b37a8a25fa2 | treesnapshot (tar.gz zip) |
| parent | 46e2ed748098f50e7ab731698723a61a4f648eea | commitdiff |
n_cmd_arg_parse(): FIX token error -> crash, e.g. "-RX 'bind;echo $?' -Xx"..
The token segmentation will not push "the empty token following
the bind command", but it will correctly split after the control
operator that the semicolon is.
The problem is that no check is performed in this case to see
whether we would have required some output, or whether the current
argument is "optional". In this case it is not, and `bind' will
blindly access the first argument, but which does not exist.
This bug exists for quite some time already, grr!
It is likely a leftover from the times when we did not support any
control operator, and when we implemented semicolon, i failed to
update all necessary places. This was one of them.
The token segmentation will not push "the empty token following
the bind command", but it will correctly split after the control
operator that the semicolon is.
The problem is that no check is performed in this case to see
whether we would have required some output, or whether the current
argument is "optional". In this case it is not, and `bind' will
blindly access the first argument, but which does not exist.
This bug exists for quite some time already, grr!
It is likely a leftover from the times when we did not support any
control operator, and when we implemented semicolon, i failed to
update all necessary places. This was one of them.
| cmd-tab.c | diffblobblamehistory |