tag 303622 + pending
tag 303339 + pending
thanks

On Thu, Apr 07, 2005 at 10:43:47PM +1000, Richard Jones wrote:
> So firstly it looks like I've been fairly boneheaded here.  that
> value of 1000 in blocks = 1000*1024 should never have been there,
> that should have been calculated from quoteMbTotal (after
> appropriate parsing into an int, the string might include the MB tag
> which will need stripping (the same way as the quotaMbUsed is
> stripped should work if it does have the {space}MB tag). This
> assumes that the libgmail.py code still calculates the correct
> values for quotainfo.

I implemented this change ("not hardcoding the gmail quota in
gmailfs"), and that effectively fixed what df reports as the total
size for a gmailfs virtual drive.

On the other hand, the problem of df reporting 0 as the usage
percentage was more a bit more involved; I had to make a change in the
python-fuse CVS repository (I actually maintain it upstream, on top of
my responsibility for it as a Debian package): it appears that the
statfs stub in _fusemodule.c wasn't dealing whatsoever with
statfs.f_bavail. This value is actually what df uses to figure out how
much space is free, whereas statfs.f_bfree is used to determine how
much space is available *to the superuser only*).

After making this change, I fixed the statfs call in gmailfs as well
(setting statfs.f_bavail to statfs.f_bfree, since I'm not sure blocks
reserved for the superuser make sense in a gmailfs context).

As soon as both the new python-fuse and gmailfs hit Debian, I'll send
you the gmailfs patch, but you'll have to make it clear that this
change *will break* if not used with the latest python-fuse from
CVS...

> Thanks again for maintaining the package, it means most of the gmailfs
> questions I get are now related to supporting the windows "Gmail
> Drive" which I have nothing to do with :-)

No problem, I'm glad my work for Debian makes your life easier :)

Cheers,

--Seb


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Reply via email to