large bucket regression

[David Nolden]

Am Dienstag 15 Dezember 2009 00:47:40 schrieb Milian Wolff:
> On Tuesday 15 December 2009 00:39:54 Milian Wolff wrote:
> > On Tuesday 15 December 2009 00:35:29 Milian Wolff wrote:
> > > Hey all, esp. David:
> > >
> > > I get this message reproducably when trying to parse kdelibs on my
> > > desktop machine. That used to work what, 2 weeks ago or something? Also
> > > afaik one of the Eigen developers reported something similar in our
> > > channel.
> > >
> > > ASSERT: "bucketPtr->largestFreeSize()" in file
> > > /home/milian/projects/kde4/kdevplatform/language/duchain/repositories/i
> > >te mr epository.h, line 1398
> > >
> > > Any idea what the issue might be? KDelibs getting bigger? KDevelop
> > >  requiering more resources?
> > >
> > > To my setup:
> > > - harddisks have some GB left
> > > - 2GB Ram (lots of other apps e.g. firefox, amarok, quassel, X also
> > >  requiring some of it)
> > > - I run a kdevelop session with only kdelibs in it
> >
> > Here's a top-output I have when I hit the assertion:
> >
> > http://mwolff.pastebin.com/fae4b973
> >
> > Note: I even close firefox and imo there should be space left. Maybe it's
> >  the KDevVarLengthArray always _duplicating_ its size (afaik?)? We really
> >  should use a memory pool here! Should also be faster (no realloc
> >  required).
> 
> Also killed amarok, no change. New top output at the assert:
> http://mwolff.pastebin.com/f5baddf23
> 
Don't worry, the amount of free memory or disk-space doesn't matter at all 
here.

It's good that you can reproduce it, as this will allow us to fix it (I 
couldn't reproduce, and there is also tests that tests testing large buckets, 
all work without a problem). Can you find out which commit led to this 
regression? The main candidates are r1061048 and r1060781.

Gretings, David