On Tue, Sep 24, 2002 at 09:21:43AM +0200, Corinna Vinschen wrote: >On Tue, Sep 24, 2002 at 01:50:54AM -0500, Steve O wrote: >> Hi, >> A week or so ago, I found that rxvt hangs could be fixed by making >> the tty pipe non-blocking. As I look into this more, I discovered >> that fhandler_pty_master::doecho uses the slave's side of the >> pipe to accomplish echoing characters to the terminal. A clever >> and simple solution, but it causes a deadlock when the slave's >> write pipe is full. >> >> The solution presented in this patch is to use a buffer to store the >> echo characters, similar to the readahead buffer. I wanted to use >> the readahead buffer, but convinced myself that the data was >> different enough that folding it in would create more complex code >> than necessary. >> >> Fixing this deadlock revealed a deadlock in the >> fhandler_pty_master::accept_input code which will be the subject of >> my next patch. > >This looks interesting but is actually missing a ChangeLog entry.
I don't think we have an assignment on file for Steve, either. (obligatory mumblings about legal crap deleted) cgf
