On Wed, Sep 09, 2009 at 12:10:15PM +0200, Hynek Schlawack wrote:
> 
> 2009/9/9 Wichert Akkerman <[email protected]>:
> 
> >> JFTR: It also happens when I merely _resize_ the windows containing
> >> the "GNU screen"...am I really the only one who encountered this
> >> problem?
> > Sounds like your paster needs to ignore a few extra signals (such as
> > SIGWINCH). I've never seen that happen though.
> 
> I've just made even more observations: It's not screen's fault.
> Resizing a terminal _always_ results in a crash by paster. I've tried
> several ones (xterm, GNOME-Terminal, Konsole) and all crash paster. I
> rarely resitze terminals but I often re-attach them, and that seems to
> cause similar signals.
> 
> As it only happens after DB access, the suspects are SQLAlchemy and/or
> cx_Oracle. However Django a test project in Django with much of the
> SQLAlchemy functionality doesn't exhibit this behaviour.

The exception is happening in

  File "/usr/local/lib/python2.6/SocketServer.py", line 262, in handle_request

This reminds me of http://bugs.python.org/issue1068268: Python's
subprocess module had issues with EINTR and was eventually fixed to wrap
select, read and write calls and retry them.

It could be that SocketServer.py needs a similar fix.  Or, perhaps, this
should be fixed once and for all in the select built-in module (as well
as read/write methods of file objects, I suppose)?

I wish someone well-versed in Unix system call semantics took interest
in this issue...

Marius Gedminas
-- 
BYTE editors are people who separate the wheat from the chaff, and then
carefully print the chaff.

Attachment: signature.asc
Description: Digital signature

Reply via email to