On Aug 2, 2009, at 7:05 AM, Kostik Belousov wrote: > On Sun, Aug 02, 2009 at 02:31:08PM +0200, Ed Schouten wrote: >> Hi Kostik, >> >> * Kostik Belousov <kostikbel_at_gmail.com> wrote: >>> I run a screen(1), where I tried to copy large portion of output and >>> paste it into vi. This resulted in the loss of the characters at >>> random >>> points inside the pasted text. >> >> I already took some time to investigate the issue. I have attached a >> patch that should already improve the situation: >> >> - write() on a pseudo-terminal master also accounted the data that >> was >> read into the kernel, but couldn't be passed to the TTY (which is >> likely to happen in non-blocking mode). >> >> - There was also a small unrelated issue; input on a TTY which has >> been >> configured in block (bypass) mode wouldn't set the input high water >> mark. >> >> For some reason, the data loss doesn't occur when SSHing to myself >> multiple times, but still causes screen(1) to drop some bytes later >> on. >> >> Even though it's always very easy to blame other applications, I >> suspect >> this may be because I reduced the input buffer size from 8 KB to 2 KB >> per pseudo-terminal. Maybe screen(1) can't deal with this. To be >> investigated... > At least, it is an improvement for me. Patch looks good. > Please consider this as an approval for the commit. SWEET! You may have fixed the bug that I reported (well, not officially) on #bsdports / #bsddev a while ago... someone else tested it though, and it passed for them so I thought it was just something funky with my setup. I suppose not :). I'll give it a shot too, once I have a chance. It was very easy to reproduce, but wasn't consistent for everyone, and the best way to reproduce it was to paste ~1k chars of text from Firefox to xterm (for instance).. -GarrettReceived on Mon Aug 03 2009 - 01:35:14 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:53 UTC