[Box Backup] (solved) More serious Win32 0.9f/RedHat 0.9 trouble...

[Nick Knight]

Win32 doesn't handle this; this was one of those things I compiled out -
with the intention of testing more heavily in the future as to why it
was needed!

I don't understand how this would cause problems in the store, my store
has ended up with some corruption on the large file, or so it appears I
still haven't got to the bottom of it yet.

Couple of questions

1/ If the client effectively never fires a timer - doesn't it just
continue trying to diff the file

2/ What is it supposed to do when the timer fires? I am still looking at
the code there!

Nick

-----Original Message-----
From: boxbackup-admin@fluffy.co.uk [mailto:boxbackup-admin@fluffy.co.uk]
On Behalf Of Gary
Sent: 14 April 2005 17:53
To: boxbackup@fluffy.co.uk
Subject: RE: [Box Backup] (solved) More serious Win32 0.9f/RedHat 0.9
trouble...

Hi everyone,

The bad news is that I started running into the large backup file
problem under Cygwin as well as Win32 (native). The good news is that I
think I found out what the problem is.

Namely, BackupStoreFileDiff.cpp uses setitimer() to get a signal when a
file diff should terminate, due to MaximumDiffingTime in bbackupd.conf.
Cygwin does not support ITIMER_VIRTUAL for setitimer() (returns EINVAL,
not sure how Win32 native handles this), so the signal is never
received (the error is ignored silently with a TRACE()). Effectively, a
file diff continues for as long as it needs, and possibly longer than
an SSL connection/session timeout, thus resulting in an error message
mentioned before in this thread.

To fix under Cygwin, patch BackupStoreFileDiff.cpp to use ITIMER_REAL
in setitimer(), and SIGALRM in signal().

Gary


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around=20
http://mail.yahoo.com=20
_______________________________________________
boxbackup mailing list
boxbackup@fluffy.co.uk
http://lists.warhead.org.uk/mailman/listinfo/boxbackup