Re: [Ifile-discuss] protect .idata from corruption

[Camillo Särs]

Jason Rennie wrote:
> I built an ifile data file /tmp/idata, maxed out my /tmp disk and then
> tried to insert a message.  ifile printed an error message and exited 
> without touching the existing data file.  Currently, it doesn't delete 
> the temporary data file.  Should it?  I closed bug 2955.

Did you try to create a race condition along the following lines?

1. Feed ifile a message to insert
2. Ifile creates temp file "somewhere else", i.e. on another partition, and
inserts message
3. Max out disk space on disk holding .idata
4. Ifile does whatever it does to get temp file to replace data file

If the error handling is incorrectly implemented, there is a chance that
during step 4. ifile first nukes the existing .idata file.  It would then
try to copy the temp file on top of it, but the disk wouldn't fit it.

Camillo
-- 
Camillo Särs <address@hidden>             Aim for the impossible and you
http://www.iki.fi/ged                   will achieve the improbable