On 04/10/12 22:05, Simon Helsen wrote:
ok, I see. Well, killing the jvm is not a fix of course, but yes, in
2.7.4, this crude workaround at least won't bust the data.
But the real point of the Issue is that one rogue query should not
jeopardize the entire application server, i.e. it needs to abort when
instructed to do so
And my suggestions are there:
[[
Possible improvements:
1/ Make TDB check the timeout during BGP execution (being careful it
does not incur appreciable cost inside a performance critical inner loop).
2/ Make union graph processing sensitive to the index used which means
it can do the duplicate suppression more efficiently.
]]
Both are changes to a small number of classes, nothing widespread.
Andy
