Solaris 11, python2.6 fixes in ptyprocess.py #15
Conversation
- termios.tcgetattr(fd)[6][VEOF] becomes value of 1(int), where we expect '\x04' (^D). seems to be a bug with the compilation of python2.6 that Sun provides with Solaris 11. if _is_solairs is True, and the return value is an integer, just assume its a bork and manually set 4(int)(^D). - Fix several "ValueError: zero length field name in format" errors for Python 2.6, which is what ships with Solaris 11. This caused an exception at a critical path in pty.fork(), where the parent would block on os.read of 'exec_err_pipe_read' indefinitely, as the child raised an exception trying to construct the 'tosend' value. - In spawn(), allow errno.ENXIO when calling inst.setwindowsize(), in some cases, such as spawn(['/bin/true']), the child process may have exited so quickly as to no longer be able to accept any terminal driver calls.
jquast
changed the title
jquast
changed the title
|
I thought we had deliberately dropped Python 2.6 support in Pexpect? I don't mind support Solaris if it has a sufficiently modern Python, but if even Python 2.7 (from 2010) is not supported, I don't think it's necessary to add support in Ptyprocess, which is a much newer piece of software. Relevant reading: Nick Coghlan's blog post Stop Supporting Python 2.6 (for free). |
|
It just isn't very hard to support python2.6. This document says, "and this is proving to be a hassle for you" -- It isn't for me.. Travis-CI still tests python2.6 for free. It seemed silly to fix @reynir's bug report only to have it lock up/freeze for him on the next release -- he or someone else will just open another bug later. In such cases not supporting py26 might be more of a hassle: We should at least make setup.py assert that python version >= 2.7 is used to prevent them from installing it and discovering and filing bug reports for such behavior. pexpect still works with python2.6, (and for tests, we need only replace unittest with unittest2 and we can stop backporting things like assertRaises in PexpectTestCase.py). Anyway, I'll make a separate PR for only the |
|
I agree it's not technically very difficult, but it's one more thing to think about - I have to remember which features I can't use because of 2.6. And it seems perverse to be adding such support to new packages. I would be fine with ptyprocess' setup.py checking the Python version (sorry @reynir). If installing a newer version of Python on Solaris is a hassle, then I think the best way forwards is for someone to work on making that process easier, rather than the whole ecosystem having to work around it. |
|
It's ok, I've moved my code away from python and pexpect. I understand the motivation for not supporting 2.6. |
termios.tcgetattr(fd)[6][VEOF]becomes value of1(int), where we expect'\x04'(^D). seems to be a bug with the compilation of python2.6 that Sun provides with Solaris 11. This raises TypeError whenord(1)happens when trying to determine the byte for EOF.If
_is_solarisis True, and the return value is an integer, just assume its a bork and manually seteof = 4(int)(^D).Fix several "ValueError: zero length field name in format" errors for Python 2.6, which is what ships with Solaris 11. This caused an exception at a critical path in pty.fork(), where the parent would block on os.read of 'exec_err_pipe_read' indefinitely, as the child raised an exception trying to construct the 'tosend' value (though the exception could not be seen). Test runner just blocks indefinitely without such fixes.
In spawn(), allow
errno.ENXIOwhen callinginst.setwindowsize(), in some cases, such asspawn(['/bin/true']), the child process may have exited so quickly as to no longer be able to accept any terminal driver calls. Reported by @reynir in Race condition in pexpect.spawn() on Solaris 11 pexpect#206