-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
We just had a bug report about the slowness of paramiko:
https://bugs.edge.launchpad.net/bugs/497733
Which relates to a bug I found a long time ago:
https://bugs.edge.launchpad.net/paramiko/+bug/271791
and
https://bugs.edge.launchpad.net/bugs/248522
Basically, paramiko uses PyCrypto's Randpool, which then tries to
measure time.time()'s tick size by calling it until it changes 100 times.
On some versions of Windows, this has a 15ms resolution, which means it
takes 1.5s of just sitting there calling time.time().
On my current Vista machine, time.time() seems to have a resolution of
1ms. (Note that time.clock() seems to have a resolution of 2.6ns, versus
1.0ms.) I believe I remember reading that there may be a 'set multimedia
timer resolution' function on Windows. And that it is set to 15ms by
default because of "system performance", but can be set down to 1ms.
Anyway, I was thinking that probably the cleanest way forward would be
to just patch Pycrypto on the build host, to use time.clock on windows:
http://launchpadlibrarian.net/16133025/win32_clock.patch
So that the standalone installers would then bundle the patched version.
Do people think that is a reasonable way to go forward?
John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
iEYEARECAAYFAksqlq0ACgkQJdeBCYSNAAMI5gCcDRcDPnBwcUoPh26UsXV/bh5e
DbwAoJV+qkLkng9L4RLudr3dJD4MuOUl
=TkBI
-----END PGP SIGNATURE-----
_______________________________________________
Mailing list: https://launchpad.net/~bzr-windows
Post to : bzr-windows@xxxxxxxxxxxxxxxxxxx
Unsubscribe : https://launchpad.net/~bzr-windows
More help : https://help.launchpad.net/ListHelp
