On Tue, 11 Feb 2014 20:24:02 -0600 Charles Cazabon <search-web-for-addr...@pyropus.ca> wrote:
> Celejar <cele...@gmail.com> wrote: > > > > I obviously can't ask for a fix for the corruption itself, since it > > isn't reproducible and I don't even know for sure if it's getmail's > > fault > > It almost certainly isn't; I agree with Matthias that this looks like a case > of user error -- i.e., you broke it, so you get to keep both pieces. getmail > simply cannot write a series of NULs to the file; there's no code in getmail > to do so, and the input messages won't have embedded NULs in them, so getmail > won't be passing them through either. The only way that the resulting file > ends up with such in it is if you aren't doing proper mbox locking and having > all programs use it (and agree on what type of locking, and agree on the mbox > type, i.e. mboxrd), and so, the file becomes corrupted as multiple processes > write to it simultaneously. That "series of NULs" is probably a hole in the > file resulting from one program truncating the file, and then the other > writing to a new location past the now-shorter end of the file. Okay. Hiroyuki Yamamoto, the Sylpheed dev, has confirmed that Sylph is using flock, so there's my problem: http://www.sraoss.jp/pipermail/sylpheed/2014-April/006010.html I've reported a bug against the Debian Sylpheed package, asking at least for documentation of this: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=746389 The Debian maintainer wonders: > Do you know why getmail has change locking method? If there's good reasons > maybe Sylpheed upstream can be convinced of changing this too. Comment? Thanks, Celejar -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
