Take back the cpu concession, make cpu/mem usage better than before the
memfix, and avoid more crashes.
With this I tried to crash it but couldn't, short of pausing the program
in certain places and changing the memory contents.
As a side note, Tomas, please take a look (in the initializer).  I'm not
sure why you used the redundant calls and I spent a long time analyzing the
code, wondering if perhaps there is some rare corner-case that needed
those.  If that's the case, I've my finger on the revert button.

CVS: ----------------------------------------------------------------------
CVS: Enter Log.  Lines beginning with `CVS:' are removed automatically
CVS: 
CVS: Committing in .
CVS: 
CVS: Modified Files:
CVS:  Tag: ABI-1-0-0-STABLE
CVS:    xap_UnixEncodingManager.cpp 
CVS: ----------------------------------------------------------------------


Hastily forport the whole mess.

CVS: ----------------------------------------------------------------------
CVS: Enter Log.  Lines beginning with `CVS:' are removed automatically
CVS: 
CVS: Committing in .
CVS: 
CVS: Modified Files:
CVS:    xap_UnixEncodingManager.cpp 
CVS: ----------------------------------------------------------------------



On Wed, 2003-01-22 at 19:21, Mark Gilbert wrote:
> This plugs the alias leak (again).
> The difference between this fix and the last attempts?  Not only
> does it plug the leak and reduce the bloat, it also does _NOT_ crash on
> startup!  There is a small price to pay.  In certain situations there will
> be statistically more cpu usage, but even on my P54C those dont even
> register time-wise on the profile (they total less than .01ms).  In return,
> we get a significantly reduced memory footprint.
> 
> I'll forport shortly.
> CVS: ----------------------------------------------------------------------
> CVS: Enter Log.  Lines beginning with `CVS:' are removed automatically
> CVS: 
> CVS: Committing in .
> CVS: 
> CVS: Modified Files:
> CVS:  Tag: ABI-1-0-0-STABLE
> CVS:  xap_UnixEncodingManager.cpp 
> CVS: ----------------------------------------------------------------------
> 
> Regards
> -MG
> 

Reply via email to